Machine Learning for Density Prediction and Process Development of Large Layer Thickness LPBF 304L Stainless Steel and Its Mechanical Impacts

This study addresses the challenge of balancing “high deposition efficiency with large layer thickness” and “component mechanical integrity” in Laser Powder Bed Fusion (LPBF) additive manufacturing. Using 304L stainless steel as an example, a hybrid modeling strategy combining physical mechanism models and residual machine learning was proposed, achieving accurate prediction of densification at H = 60, 90, and 120 μm (test set R² = 0.833, MAE = 0.104). Within the Doehlert matrix experimental design framework, the coupled effects of laser power, scanning speed, and scanning spacing on densification behavior, microstructure evolution, and mechanical response at different layer thicknesses were systematically analyzed. The results show that, after model-driven parameter optimization, the relative density of the 60 μm thick sample can reach 99.98%, and it achieves mechanical properties with both high strength and high ductility (tensile strength 695.5 MPa, yield strength 531 MPa, elongation after fracture 50%). The strengthening effect can be attributed to the synergistic effect of grain refinement, a high proportion of large-angle grain boundaries (HAGBs), and high-density nano-oxides. As the layer thickness increases to 120 μm, although some grain coarsening and oxide particle thinning occur, good melt pool stability and densification levels can still be maintained through process window control under the guidance of the hybrid model, ensuring that the overall material performance remains at a high level. Overall, the “mechanism-data fusion” prediction framework established in this paper provides an interpretable parameter optimization path and experimental basis for the development of large-layer-thickness LPBF processes, and offers a reference for achieving a synergistic improvement in manufacturing efficiency and component quality.