Data-driven modelling of hot workability integrating microstructure state and deformation conditions

Abstract Objectives The hot processing map serves as a fundamental tool for evaluating hot workability and optimizing processing parameters. However, the hot processing map establishes a direct relationship between hot workability and deformation conditions, leaving the mechanisms underlying hot workability evolution unclear and limiting the applicability. This paper aims to develop a method for predicting hot workability that comprehensively accounts for the effects of both deformation conditions and microstructural states. Methods Thermal deformation and microstructural characterization are carried out on a low-carbon steel to investigate evolution in hot workability within typical hot working ranges (900–1050°C, 0.01–10 s−1). A neural network model integrated with Shapley additive explanations is then employed to establish the coupling relationships among deformation conditions, microstructural states, and hot workability. Results Dislocation density and dynamic recrystallization fraction and deformation conditions exert a coupled effect on hot workability. A prediction framework for hot workability is proposed, integrating microstructure states and deformation conditions, and comprising a coupled internal state variable constitutive model and two neural network models. The complex interactions among deformation conditions, microstructure evolution, and hot workability are thoroughly demonstrated under representative conditions. Conclusion Dynamic recrystallization softening and sustained work hardening represent two distinct mechanisms for enhancing hot workability at low and high strain rates, respectively. The proposed predictive framework reveals the coupled relationships among deformation conditions, microstructural state, and hot workability.