Revealing thickness-gradient microstructure and hardness evolution in a laser-treated 2.0 GPa ultra-high strength steel

Accurate control of local properties in laser-processed press-hardened steels (PHS) is hindered by steep through-thickness thermal gradients and the generate heterogeneous phase transformation paths. This study establishes a quantitative process-thermal history-transformation regime-hardness linkage for Al-Si coated PHS2000 subjected to laser surface heat treatment. A 3D transient thermal finite element model, validated by infrared thermography, reconstructs temperature-time histories and partitions thermal cycles into subcritical (tempering), intercritical (partial austenitization), and fully austenitic regimes using and thresholds. The resulting transformation regime map elucidates the non-monotonic hardness gradients: a hardened surface layer derived from full austenitization and quenching, and a subsurface softening zone associated with intercritical exposure where diffusion-controlled tempering dominates. To enable rapid and traceable property prediction, we develop a physics-guided machine learning framework using transformation-aware thermal descriptors extracted from 450 thermal histories. Compared with a baseline empirical model relying only on process parameters, the physics-guided model, incorporating kinetic features such as cooling rate and time above , achieves significantly higher accuracy (testing = 0.89, = 23.9HV) and maintains reasonable generalization on an independent variable-speed validation set ( = 0.89, = 26.9HV). Shapley Additive Explanations (SHAP) analysis further identifies peak temperature, thermal exposure and cooling rate as the dominant drivers, rationalizing the competition between partial austenitization hardening and thermal softening. The identified process-thermal history-hardness linkage offers a physically grounded basis for optimizing laser surface treatment parameters to achieve tailored through-thickness property profiles in ultra-high-strength steels.