Using machine-learning-based process map to guide evaluation of Inconel 625 mechanical properties in laser foil printing

Inconel 625 is widely studied in powder-based additive manufacturing, but its processing characteristics and mechanical performance in foil-feedstock laser foil printing (LFP) remain largely unexplored. In this study, a gradient boosting regression (GBR)-based process map was developed for LFP of Inconel 625 using 32 single-track experiments. The GBR model achieved R 2 values of 0.859 and 0.793 and mean absolute errors of 22.25 μm and 19.83 μm for melt-pool depth and width, respectively, outperforming second- and third-order polynomial regressions in capturing nonlinear melt-pool responses and distinguishing lack-of-fusion, conduction, and keyhole regimes. Three conduction-mode conditions with target depth-to-foil thickness ratios (D/T) of approximately 1.3, 1.5, and 1.7, along with one keyhole-mode condition, were selected for multilayer validation. The conduction-mode builds showed uniform melt pools, sound interlayer bonding, ultra-low porosity below 0.05%, and excellent tensile properties, including ultimate tensile strength of 888–988 MPa, yield strength of 708–759 MPa, and elongation of 35–41%. In contrast, the keyhole-mode sample retained comparable strength but exhibited reduced ductility due to keyhole-induced gas pores, a higher high-angle grain boundary fraction, and reduced XRD peak broadening. These results demonstrate that GBR-driven process mapping can effectively guide parameter selection and enable dense, high-performance Inconel 625 components by LFP.