A probabilistic fatigue life model for the evaluation of casting defects in a nickel-based superalloy under high temperature LCF-loading

This work presents a physically based probabilistic model for LCF life of the conventional cast (CC) nickel-based superalloy MAR-M247. The model explicitly accounts for the effects of casting defects. Over 100 LCF tests were conducted at 850 °C and 950 °C on specimens with and without defects. Specimens were extracted from turbine blades that had manufacturing-related casting defects. Additionally, artificial defects were introduced into initially defect-free specimens. A multi-stage CT- and μ CT-based specimen extraction strategy was developed to precisely position casting defects within specimens. Defects were parameterized orthogonally to the loading direction based on their size, shape and orientation using SEM fractography. For all tested strain ranges, a dominant influence of defect size on fatigue life was found. The effects of defect shape and orientation were small in comparison. Significant scatter of material parameters was observed experimentally. Electron backscatter diffraction (EBSD) revealed significant variations in grain size and orientation depending on the specimen extraction position. The model accounts for this by using individual values for stress ratio R σ and Young’s modulus E . LCF fatigue life was modeled using an approach based on the cyclic crack tip opening displacement Δ C T O D . Defect size was accounted for as an initial crack length based on the Feret diameter d Feret of each defect. Additionally, an El Haddad-type threshold term was used to account for defect size dependence in the high-cycle regime. Validation against experimental data showed robust prediction accuracy across all defect types, defect sizes, specimen geometries, temperatures, and strain ranges. Model inaccuracies arising from the fracture mechanics based approximation solution were estimated analytically. For defects large compared to the specimen geometry, a tendency towards more conservative fatigue life assessment was shown. A probabilistic framework was established using Monte Carlo simulations. Material scatter was modeled using experimentally derived distributions for E and R σ . Fatigue life distributions were quantified under realistic operating temperatures and loads. The results support risk-based maintenance and optimized gas turbine operation in modern decarbonized energy systems.