Enhancing flow stress predictions in CoCrFeNiV high entropy alloy with conventional and machine learning techniques

Abstract

A machine learning technique leveraging artificial intelligence (AI) has emerged as a promising tool for expediting the exploration and design of novel high entropy alloys (HEAs) while predicting their mechanical properties at both room and elevated temperatures. In this paper, we predict the flow stress of hot-compressed CoCrFeNiV HEAs using conventional (qualitative and quantitative models) and advanced machine learning approaches across various temperature and strain rate conditions. Conventional modeling methods, including the modified Johnson-Cook (JC), modified Zerilli–Armstrong (ZA), and Arrhenius-type constitutive equations, are employed. Simultaneously, machine learning models are utilized to forecast flow stress under different hot working conditions. The performance of both conventional and machine learning models is evaluated using metrics such as coefficient of determination (R2), mean abosolute error (MAE), and root mean squared error (RMSE). The analysis reveals that the gradient boosting machine learning model shows superior prediction accuracy (with value R2 = 0.994, MAE = 7.77%, and RMSE = 9.7%) compared to conventional models and other machine learning approaches.