Data-driven optimization of laser scanning conditions in laser powder bed fusion for defect-free IN738LC components

Manufacturing defect-free superalloy components using laser powder bed fusion (L-PBF) remains a challenging and urgent issue due to the high susceptibility of the process to microcracking. While previous studies have investigated how individual laser scanning parameters influence microcracking behavior in superalloy components, a comprehensive process window for crack density has yet to be developed. Moreover, the relationship between laser scanning conditions and crack density is often treated as a black-box model, without explicitly considering the underlying fundamental mechanisms. Thus, the purpose of this study is twofold: first, to establish a comprehensive process window for crack density in order to fabricate defect-free IN738LC components by L-PBF; and second, to elucidate the fundamental mechanisms linking laser scanning conditions to solidification cracking through empirical causal analysis based on microstructural features. To this end, more than 100 sets of laser scanning conditions were investigated, and the optimal conditions were found to minimize the defect ratio and crack density to <0.060 % and 0.005 mm/mm², respectively. The suitability was further validated by fabricating turbine-blade shaped parts without internal defects. Then, microstructural features for all samples were extracted using electron backscatter diffraction, and the resulting dataset was used to develop regression models for predicting crack density. The multiple linear regression and support vector regression models revealed that two common key microstructural features—grain refinement and the alignment of 〈001〉 to the building direction—play a primary role in suppressing microcracking. On the other hand, the findings also imply that incorporating additional metallurgical and mechanical features may be essential for enhancing the predictive performance.