Physics-Guided Decision-Support Framework for Melt Pool Prediction and Process Stability in Laser Powder Bed Fusion of Nitinol

Abstract

Powder bed fusion–laser beam (PBF-LB) of nickel–titanium (NiTi) has attracted increasing interest in aerospace, biomedical, and energy applications owing to its shape memory and superelastic properties, combined with the capability to fabricate complex geometries. However, the strong sensitivity of NiTi to thermal history and process variability makes predictive modeling and process parameter selection challenging. In this work, a physics-guided decision-support framework is developed for melt pool prediction and stability assessment during the PBF-LB processing of NiTi. A high-fidelity thermal finite element model incorporating CALPHAD-derived, temperature-dependent material properties was calibrated using a subset of experimental measurements and independently validated against additional experimental melt pool data. The calibrated model demonstrated good agreement with experiments, yielding mean absolute percentage errors of 4.22% and 5.63% for melt pool width and depth, respectively, on the validation dataset. A multi-output random forest surrogate trained on the validated simulation dataset enabled rapid prediction of melt pool geometric features, achieving test-set R2 values exceeding 0.95, together with low MAE and RMSE values, while five-fold cross-validation confirmed robust predictive performance. The proposed framework integrates surrogate predictions with physics-based melt pool stability criteria to rapidly identify physically feasible processing conditions, thereby providing a computationally efficient foundation for future supervisory process control strategies.