Comparison and evaluation of different constitutive models for predicting the hot deformation behavior of Mg-Gd-Y-Zr alloy

Abstract

The popular constitutive models used in the field of hot forming of magnesium alloys can be divided into phenomenological models, machine learning models, and internal state variables (ISV) models based on physical mechanisms. Currently, there is a lack of comparison and evaluation regarding the suitability of different types of models. In this study, Mg-Gd-Y-Zr alloy is taken as the research object. The hot deformation behavior of the alloy was studied systematically. Subsequently, Arrhenius model with strain compensation, artificial neural network (ANN) model, and ISV model involving dynamic recrystallization (DRX), dislocation density and grain size evolution were established. ANN model demonstrates a higher level of accuracy in fitting the original stress-strain curves compared to both ISV model and modified Arrhenius model, but ANN model is not suitable for predicting the experimental results outside of the initial database. ISV model considers the impact of microstructure evolution history on stress, making it highly effective in reflecting the mechanical responses under complex loading condition. The established ISV model is embedded in the ABAQUS software, which shows good ability in calculating the mechanical response, dimension, and microstructure evolution information of the component during hot forming.