Comprehensive Assessment of Constitutive Models for Precise Flow Stress Prediction in Aluminum Matrix Composites under Thermomechanical Loading at Elevated Temperatures

Optimizing hot-working process requires an accurate prediction of the flow stress behavior of aluminum matrix composites (AMCs) at high temperatures. In this study, hot-compression tests were conducted on 15% SiCp/AA2024 composites to evaluate three constitutive models: Arrhenius, Double Multiple Nonlinear Regression (DMNR), and Modified Johnson-Cook (mJ-C). Experiments were performed on a Gleeble−3500 simulator at temperatures ranging from 673 to 753 K, strain rates between 0.01 and 1 s⁻¹, and true strains up to 0.7. Based on the statistical indicators, including correlation coefficient (R), average absolute relative error (AARE), and root-mean-square error (RMSE), the DMNR model demonstrated the highest predictive accuracy (R = 0.99467, AARE = 1.8080%, and RMSE = 1.7968 MPa) outperforming both the mJ-C and Arrhenius models. A key contribution of this work is the direct construction of hot-processing maps using the DMNR model, which has not been reported previously. Using this model, the strain-rate sensitivity (m), the strain-hardening exponent (n), and s", enabling clear identification of stable and unstable deformation regions. These maps highlight processing windows with high energy-dissipation efficiency and provide a practical basis for process optimization. Furthermore, the microstructural observations confirmed that DMNR-predicted optimal zones correspond to regions with fine, recrystallized grains, supporting the reliability and applicability of the DMNR-based processing maps.