Machine learning approach to predict new multiphase high entropy alloys

Abstract

High entropy alloys with multi-principal elements have interested the research community due to the promising properties and tunable microstructure. In the current study, the multiphase alloy system with a mixture of solid solution and intermetallic (SS+IM) was predicted using a machine learning approach with a data set of 636 alloys. The Algorithms used are Logistic Regression, Decision Tree, Support Vector Machine (SVM) classifier, Random Forest, Gradient Boosting Classifier, and Artificial Neural Network (ANN). ANN has shown the best accuracy of more than 80% for the test data. The new alloys were prepared and characterized to verify the prediction and it is found that ANN is having more accurate prediction in the studied alloy system. Statistical analysis of the established data set reveals an overlapping boundary between the design parameters that hinders the successful prediction. Experimental data confirms the formation of new multiphase alloys.