Insights on phase formation from thermodynamic calculations and machine learning of 2436 experimentally measured high entropy alloys
Insights on phase formation from thermodynamic calculations and machine learning of 2436 experimentally measured high entropy alloys
复制标题
从 2436 种实验测量的高熵合金的热力学计算和机器学习中了解相形成
DOI：
10.1016/j.jallcom.2022.165173
复制
发表时间：
2022
影响因子：
6.2
通讯作者：
Zhao, Ji-Cheng
中科院分区：
文献类型：
--
作者：
Wang, Chuangye;Zhong, Wei;Zhao, Ji-Cheng
摘要采用相图计算（CALPHAD）和机器学习（ML）方法分析了2436种由Al、Co、Cr、Cu、Fe、Mn和Ni组成的高熵合金（HEA）成分的相形成。结果表明，CALPHAD在预测1761固溶组合物的BCC/B2和FCC相形成方面具有良好的能力，不包括含有非晶相（AM）或/和金属间化合物（IM）的HEAs。采用多个参数系统考察了相选择规律，发现价电子浓度(VEC)< 6.87和VEC> 9.16分别是形成单相BCC/B2和FCC的条件；而CALPHAD可以100%准确地预测这一点。CALPHAD预测和实验观察均表明，随着元素间原子尺寸差异的增大，BCC/B2合金的形成比FCC合金多。采用决策树（DT）、k近邻（KNN）、支持向量机（SVM）和人工神经网络（ANN）四种机器学习算法，研究了两个不同数据集的相选择规则，一个数据集包含1761个不含AM和/或IM相的固溶体HEA，另一个数据集包含所有2436个HEA成分。对ML模型进行交叉验证（CV）， DT、KNN、SVM和ANN预测BCC/B2、BCC/B2+ FCC和FCC的CV准确率分别为90.4%、94.1%、93.8%、89.7%；DT、KNN、SVM和ANN预测SS、AM、SS+ AM和IM的准确率分别为92.9%、96.3%、96.9%、92.3%。用训练好的神经网络模型对66种具有SS结构的大块合金进行了预测，预测精度达到80.3%。VEC是BCC/B2、BCC/B2+ FCC和FCC相预测中最重要的参数。电负性差和FCC-BCC-index （FBI）是决定SS、AM、SS+ AM和IM形成的两个主要特征。在Δ H mix-vs-VEC图中发现了一条分离线Δ H mix= 29× VEC−247，用于预测单相BCC/B2或FCC的形成，准确度为96.2% （Δ H mix=混合焓）。这些见解将对设计具有目标晶体结构的HEAs非常有价值。
Abstract Both CALPHAD (CALculation of PHAse Diagrams) and machine learning (ML) approaches were employed to analyze the phase formation in 2436 experimentally measured high entropy alloy (HEA) compositions consisting of various quinary mixtures of Al, Co, Cr, Cu, Fe, Mn, and Ni. CALPHAD was found to have good capabilities in predicting the BCC/B2 and FCC phase formation for the 1761 solid-solution-only compositions, excluding HEAs containing an amorphous phase (AM) or/and intermetallic compound (IM). Phase selection rules were examined systematically using several parameters and it was revealed that valence electron concentration (VEC)< 6.87 and VEC> 9.16 are the conditions for the formation of single-phase BCC/B2 and FCC, respectively; and CALPHAD could predict this with essentially 100% accuracy. Both CALPHAD predictions and experimental observations show that more BCC/B2 alloys are formed over FCC alloys as the atomic size difference between the elements increases. Four ML algorithms, decision tree (DT), k-nearest neighbor (KNN), support vector machine (SVM), and artificial neural network (ANN), were employed to study the phase selection rules for two different datasets, one consisting of 1761 solid-solution (SS) HEAs without AM and/or IM phases, and the other set consisting of all the 2436 HEA compositions. Cross validation (CV) was performed to optimize the ML models and the CV accuracies are found to be 90.4%, 94.1%, 93.8%, 89.7% for DT, KNN, SVM, and ANN respectively in predicting the formation of BCC/B2, BCC/B2+ FCC, and FCC; and 92.9%, 96.3%, 96.9%, 92.3% for DT, KNN, SVM, and ANN respectively in predicting SS, AM, SS+ AM, and IM phases. Sixty-six experimental bulk alloys with SS structures are predicted with the trained ANN model, and the accuracy reaches 80.3%. VEC was found to be most important parameter in phase prediction for BCC/B2, BCC/B2+ FCC, and FCC phases. Electronegativity difference and FCC-BCC-index (FBI) are the two dominating features in determining the formation of SS, AM, SS+ AM, and IM. A separation line Δ H mix= 29× VEC− 247 was found in the Δ H mix-vs-VEC plot to predict the formation of single-phase BCC/B2 or FCC with a 96.2% accuracy (Δ H mix= mixing enthalpy). These insights will be very valuable for designing HEAs with targeted crystal structures.