The rapid development of additive manufacturing (AM) offers unprecedented design freedom for high-performance aluminum alloys, yet process optimization remains challenging due to the complex interdependencies between processing parameters and mechanical properties. Here, we present an integrated machine learning (ML) framework coupling predictive modeling with genetic algorithm (GA) optimization to accelerate the discovery of optimal processing conditions for laser powder bed fusion (LPBF) of aluminum alloys based on a 2024 Al alloy with Ti and B addition. Leveraging a comprehensive experimental data set, this work established a high-accuracy predictive model (Light Gradient Boosting Machine, LGBM) that predicts mechanical properties from composition and process parameters. The GA-guided optimization identified optimal parameters that yielded exceptional mechanical performance with a tensile strength of 505 MPa and a fracture elongation of 8.5%, surpassing all entries in the training data set. Microstructural analysis revealed that in situ-formed TiB2 nanoparticles promoted heterogeneous nucleation, resulting in ultrafine equiaxed grains. The strengthening mechanisms were attributed to grain boundary strengthening from the refined microstructure and dislocation strengthening arising from the high density of geometrically necessary dislocations (GNDs) generated by thermal mismatch under rapid solidification conditions, as well as dislocation bypassing of second-phase particles. This work demonstrates that ML-driven optimization can effectively navigate the high-dimensional parameter space of AM, offering a process-property modeling workflow for accelerating the design of high-performance, customized Al alloys.
Wang et al. (Sat,) studied this question.