机器学习指导相和硬度可控的AlCoCrCuFeNi系高熵合金设计

Machine learning guided phase and hardness controlled AlCoCrCuFeNi high-entropy alloy design

采用机器学习辅助高熵合金设计,致力于解决传统试错实验方法时间周期长、成本高的问题。以经典的AlCoCrCuFeNi系高熵合金为研究对象,采用机器学习方法,分别构建高熵合金的相结构预测模型和硬度预测模型。其中支持向量机模型(SVM)在两个任务中均有最好的训练表现,最佳的相分类准确率达0.944,硬度预测模型的均方根误差为56.065HV。进一步串联两种机器学习模型,基于样本数据集上下限的成分空间,对AlCoCrCuFeNi系高熵合金同时进行相和硬度的高效预测和筛选,实现新型合金成分的快速设计。实验验证5种新合金符合相预测结果,测试硬度与预测硬度值的RMSE为12.58HV,表明建立的机器学习模型实现对高熵合金相和硬度的高效预测。

Machine learning(ML)assisted high-entropy alloys(HEA)design is dedicated to solving the problem of long period and high cost of designing by traditional trial and error experimental methods.The classic AlCoCrCuFeNi HEA was taken as the research object.The phase structure prediction model and hardness prediction model were established respectively.The support vector machine(SVM)models have the best training performance in both tasks.The best phase classification accuracy is 0.944,and the root mean square error(RMSE)of the hardness regression model is 56.065HV.The two ML models are further connected in series.Based on the upper and lower limits of the data set,the high-throughput prediction and selection of phases and hardness of AlCoCrCuFeNi HEA are carried out simultaneously,thus realizing the efficient composition design of the new alloy.The experimental results show that the five new alloys are in accord with the predicted results,and the RMSE is 12.58HV.It shows that the ML models can predict the phase and hardness of HEA efficiently and accurately.