A Probabilistic Fatigue Framework to Enable Location-Specific Lifing for Critical Thermo-mechanical Engineering Applications

The present paper describes a probabilistic framework to predict the fatigue life and failure mode under various thermo-mechanical loading conditions. Specifically, inclusion- and matrix-driven competing failure modes are examined within nickel-based superalloys. The critical accumulated plastic strain energy density (APSED) is employed as a unified metric to predict fatigue crack initiation in metals, which is favorable due to the usage of a single unknown parameter and its capability to predict failure across loading conditions and failure modes. In this research, we characterize the temperature-dependent variation of the critical APSED using a Bayesian inference framework and predict the competing failure modes in a coarse grain variant of RR1000 with varying strain range and temperature. The critical APSED appears to decrease along a vertically reflected sigmoidal curve with increasing temperature. Further, (a) the prediction of a failure mode, (b) failure mode associated with the minimum life, and (c) the change in the location associated with the matrix-driven failure mode with increasing temperature and decreasing strain range are consistent with the experimentally observed trends in RR1000, as well as other Nickel-based superalloys, documented in the literature. Finally, for each simulated loading condition, the uncertainty in the fatigue life is quantified as a prediction interval computed based on a 95%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$95\%$$\end{document} confidence level of the critical APSED and the computed APSED from simulations. The overall framework provides a promising step towards microstructural-based fatigue life determination of components and enables a location-specific lifing approach.