Comparative analysis of modified Johnson-Cook model and artificial neural network for flow stress prediction in BN-reinforced AZ80 magnesium composite.

Abstract

Boron nitride (BN), renowned for its exceptional optoelectrical properties, mechanical robustness, and thermal stability, has emerged as a promising two-dimensional material. Reinforcing AZ80 magnesium alloy with BN can significantly enhance its mechanical properties. To investigate and predict this enhancement during hot deformation, we introduce two independent modeling approaches a modified Johnson-Cook constitutive model and an artificial neural network (ANN). These models aim to capture both linear and nonlinear deformation characteristics. Hot compression tests conducted across various temperatures and strain rates provided a comprehensive dataset for model validation. The MJCC model, accounting for strain rate and temperature effects, achieved a correlation coefficientRof 0.96 and an average absolute relative error (AARE) of 6.28%. In contrast, the ANN, trained on experimental data, improved the correlation coefficient toRof 0.99 and reduced the AARE to below 1.5%, significantly enhancing predictive accuracy. These results indicate that while the modified J-C model provides reliable predictions under moderate conditions, the ANN more effectively captures complex behaviors under extreme deformation conditions. By comparing these modeling approaches, our study offers valuable insights for accurately predicting the rheological behavior of BN-reinforced AZ80 magnesium composite, aiding process optimization in industrial applications.