Machine learning assisted laser powder bed fusion process optimization of CM247LC: crack mitigation and strength–ductility enhancement

Abstract Cracking poses a major challenge in laser powder bed fusion (L-PBF) of nickel-based superalloys and is highly sensitive to laser scanning parameters. Traditional trial-and-error parameter optimization process is costly and inefficient. This study presents a machine learning (ML)-assisted strategy to optimize L-PBF parameters for the crack-prone superalloy CM247LC, aiming to suppress cracking and improve the strength–ductility trade-off. An orthogonal experimental dataset linking processing parameters to crack density was first established. An ML-based crack prediction model was then developed and enhanced through data augmentation and Bayesian optimization, increasing the prediction accuracy (R 2 from <0.55 to >0.85) and global search capability. Within only two iterations, the crack density was reduced by 99% compared to the best orthogonal result, achieving nearly crack-free samples. Microstructural analysis indicated that the improved cracking resistance under the ML-optimized parameters is attributed to weakened Hf/C segregation at the grain boundaries, a lower fraction of high-angle grain boundaries, and reduced residual stress. At 900 °C, the L-PBF-processed alloy demonstrated a 23% increase in ultimate tensile strength ((937 ± 8) MPa), a 35% increase in yield strength ((809 ± 5) MPa), and comparable elongation ((10.8 ± 0.4)%) over cast CM247LC, exhibiting superior mechanical performance among the currently reported L-PBF-processed CM247LC. The proposed ML-driven optimization framework offers an efficient and economical route for mitigating cracking in L-PBF-fabricated crack-sensitive alloys.