A New Extreme-Value Probabilistic Framework for Predicting Fatigue Crack Initiation Life of Notched Components

Abstract

Traditional deterministic methods for predicting the fatigue life of notched components require a number of approximations based on heuristics and phenomenological data rather than solid theoretical underpinning and still yield unsatisfactory and inconsistent results when applied to complex components under service loads. Microstructural inhomogeneities in the materials are still an important issue, but are not explicitly accounted for in the traditional deterministic methods. Recent developments in computational crystal plasticity and microstructure-scale modeling have provided deeper understanding of the complex correlations between properties and structures and further indicate the limitations of conventional fatigue life prediction approaches. These modeling approaches have the potential to substantially reduce the need for costly large scale experimental programs to determine scatter in fatigue, for example. At present, however, there is a lack of simulation-based strategy for considering interactive effects of stress/strain field gradients at the notch-root and microstructure-scale behavior in predicting notch-root fatigue crack initiation. In this paper, the distribution of a shear-based fatigue indicator parameter computed within a well-defined fatigue damage process zone at the notch are used along with a novel probabilistic mesomechanics approach to obtain the probability distribution of fatigue crack initiation of notched components, thus extending fatigue life prediction to explicitly incorporate microstructure sensitivity via probabilistic arguments. The new probabilistic framework presented in this paper takes into account the complete plastic shear strain field around the notch root and also links the variation in the materials microstructure and associated slip activations to observable scatter in fatigue strength of the notched component. The use of such probabilistic approach can be beneficial as it avoids conservatism that may result from the use of deterministic approach for fatigue life prediction.