Machine-learning-assisted optimization of constitutive modeling and microstructural insights into the hot deformation of L12-strengthened Co33Cr23Ni34Al5Ti5 chemically complex alloys

Abstract

To tackle the strength-ductility trade-off in chemically complex alloys (CCAs), multi-step thermo-mechanical processing, particularly hot deformation, is a key strategy for optimizing microstructures. Hot deformation controls dynamic recrystallization (DRX) behaviors, enabling the regulation of heterogeneous structures and precipitated phases. This study investigates the influence of temperature and strain rate on the deformation behavior and microstructural evolution of L1₂-strengthened Co₃₃Cr₂₃Ni₃₄Al₅Ti₅ CCAs. The dynamic materials model-derived processing map determines 1313 K and 0.001 s⁻¹ as the optimal processing window, validated by microstructural observations. Moreover, an advanced eXtreme Gradient Boosting (XGBoost)-assisted machine learning (ML) model is developed, demonstrating superior predictive accuracy compared to traditional constitutive models. At low-strain rates, increasing temperature induces the transition from the dominant DRX mechanism to discontinuous DRX (DDRX) to a coupled DDRX and continuous DRX (CDRX) regime. Conversely, higher strain rates at elevated temperatures weaken CDRX. Additionally, the presence of L1₂ nanoprecipitates effectively controls recrystallized grain growth by pinning dislocations and impeding subgrain boundary movement, leading to microstructure refinement. These findings offer critical insights for optimizing thermo-mechanical processing, providing a pathway to design advanced L1₂-strengthened CCAs with tailored microstructures.