Evaluation of Hot Flow Behavior Models in FeCoCrNiAl0.1 High-Entropy Alloys by Modified Johnson-Cook Model, Modified Zerilli-Armstrong Model and GA-BP Neural Network

Abstract

The hot flow behavior of FeCoCrNiAl_0.1 high-entropy alloys (HEAs) was investigated using a Gleeble−3500 thermal simulation test machine under hot deformation conditions at 950-1100°C/0.001-1 s^-1. A modified Zerilli-Armstrong (Z-A) model, a modified Johnson-Cook (J-C) model, and a GA-BP neural network were devised to predict the hot flow behavior. Model accuracy was evaluated using correlation coefficients (R²) and mean absolute relative error (AARE). Additionally, electron backscatter diffraction (EBSD) analysis was employed to examine microstructural evolution of the studied alloy after hot compression experiments. The results indicate that the modified Z-A model yielded R² of 0.9354 and AARE of 7.52%, while the modified J-C model attained R² of 0.9435 and AARE of 6.38%. Notably, GA-BP neural network exhibits the highest accuracy, with R² reaching 0.9969 and AARE of 2.66%. Microstructural analysis revealed that many fine recrystallized grains were formed at grain boundaries and within grain interiors. Continuous dynamic recrystallization (CDRX) and discontinuous dynamic recrystallization (DDRX) occurred simultaneously during hot deformation.