Bayesian Hierarchical Fatigue Scattering Model for Laser-Fused Metal Components

Laser Powder Bed Fusion (LPBF) enables the fabrication of complex metallic components with high precision and flexibility. However, LPBF-manufactured materials often exhibit substantial scatter in fatigue behavior due to process-induced defects and microstructural variability. This scatter, combined with limited testing budgets and frequent runouts, leads to high uncertainty in S–N curve estimation, particularly in estimating the endurance limit. When analyzing multiple manufacturing conditions in LPBF, fitting each condition independently can be unreliable with sparse data, while pooling all data into a single curve can obscure process-specific effects. To address this challenge, this work proposes an integrated framework that combines a Bayesian hierarchical censored S–N model with Fisher information matrix (FIM)-based D-optimal experimental design. The hierarchical model shares information across manufactured samples while preserving sample-specific S–N curves and endurance limits. The D-optimal design selects stress levels for testing that emphasize high-information regions. Together, these components improve uncertainty quantification and support more efficient fatigue testing.