Design of High Entropy Superalloy FeNiCrAlCu using Computational Thermodynamic and Machine Learning: Effect of Alloying Compositions and Temperatures on the Stacking Fault Energy

Abstract

High entropy superalloys (HESA) have great potential to replace superalloys with promising properties extensively developed to improve performance, resource sustainability, and cost efficiency in high-temperature applications. This study focuses on Fe-based HESA and their stacking fault energy (SFE), a critical parameter influencing deformation mechanism and creep resistance. This development is economically cheaper since it utilizes Fe rather than Ni as the alloy base, which has been widely developed. We propose a novel approach for predicting SFE using big data analysis leveraging machine learning and computational thermodynamics. The calculated SFE as a function of compositions and temperature becomes the database for the machine learning model. We employ a deep learning neural network model to achieve an impressive 0.008 Root Mean Squared Error (RMSE) predicting SFE values and classes. The composition of the high entropy superalloy is designed to lower the SFE, which promotes the formation of stacking faults and twin boundaries, resulting in high strength and creep resistance at high temperatures. Our research establishes an optimal design guide for achieving desired SFE: Ni (9-15 at%), Cr (15-36 at%), Al (5-22.75 at%), Cu (9-22.75 at%), and Fe (22.75-40 at%). Fe can be increased until 40 at.% with 15 at.% Ni, or Ni can be reduced until 9 at.% with a lower Fe of 22.75 at.%.