Machine learning guided constitutive model and processing map for Fe2Ni2CrAl1.2 multi-principle element alloys

• Published in: Journal of materials research and technology 2024-03, Vol.29, p.353-363
• Main Authors: Qiao, Ling, Inoue, Junya, Zhu, Jingchuan
• Format: Article
• Language: English
• Subjects: Fe2Ni2CrAl1.2 MEAs; Hot deformation behavior; Hot processing map; Machine learning constitutive model
• ISSN: 2238-7854
• DOI: 10.1016/j.jmrt.2024.01.119

The present work provides a systematical investigation on the hot deformation behavior of Fe2Ni2CrAl1.2 multi-principle element alloys (MPEAs). Hot compression tests were carried out at various temperatures ranging from 800 °C to 1100 °C at different strain rates from 0.001 s-1 to 1 s-1. The stress–strain curves obtained under different processing conditions were used to develop the constitutive model for the alloy. The advanced machine learning (ML) models including Support Vector Regression (SVR), Artificial Neural Network (ANN), Generalized Regression Neural Network (GRNN) and Random Forest (RF) were trained to predict the flow behavior of Fe2Ni2CrAl1.2 MPEAs. The predictive capability of ML algorithms were estimated, the SVR model performed better among the developed four algorithms. Then Particle Swarm Optimization (PSO) was imposed to SVR model to further enhance its prediction accuracy. The developed PSO-SVR model achieved an average testing R2 of 0.9819, as well as low RMSE and MAPE values, demonstrating its strong predictive performance. Then the PSO-SVR model was applied to generate the flow curves at various temperature and strain rate for the development of the hot processing maps. The flow instability and the optimum processing conditions were identified, indicating that instability (ζ