Accelerating lPBF process optimisation for NiTi shape memory alloys with enhanced and controllable properties through machine learning and multi-objective methods

NiTi shape memory alloys (SMAs) prepared by the laser powder bed fusion (LPBF) technology have demonstrated promise in aerospace and medical applications. Nevertheless, ensuring repeatability and customised design in printed parts remains challenging. This paper addressed this challenge by introducing a machine learning model that effectively predicted the performance of NiTi SMAs across diverse LPBF processing and equipment conditions. Trained on a dataset of 195 entries from 23 publications, the model accurately predicted critical metrics, including density, ultimate tensile strength, elongation, and thermal hysteresis. Validation using data from eight experimental groups confirmed its reliability and generalisation capability. Multi-objective optimisation identified processes yielding synergistic improvements, achieving a tensile strength of 783 $\pm$ ± 8 MPa, an elongation of 13.7 $\pm$ ± 0.8% and a low hysteresis of 15.1 K. This study also discussed strategic applications of the model for LPBF process optimisation and proposed a method for constructing tailored LPBF process maps for specific NiTi alloy performance attributes.