Probabilistic Formulation of the Cyclic Void Growth Model to Predict Ultralow Cycle Fatigue in Structural Steel

Abstract

The previously developed cyclic void growth model (CVGM) has been demonstrated to accurately simulate ductile fracture initiation under monotonic and ultralow cycle fatigue loading for a variety of steel materials and geometric configurations. Prediction of ductile fracture initiation involves significant uncertainty, particularly where there is high variability in the material (e.g., welded connections) subjected to irregular cyclic loading. The reliability of the model predictions is improved through a probabilistic formulation based on maximum likelihood parameter estimation. The probabilistic formulation, which incorporates information from both the failure and nonfailure loading cycles, has the following features: (1) the calibration of model parameters provides the maximum likelihood of agreement for a given set of cyclic fracture observations, and (2) fracture predictions are provided in a probabilistic sense by generating a distribution of the expected instant of fracture. The benefit of the approach is twofold. First, it eliminates an inconsistency that is inherent in the deterministic calibration procedure, as proposed in the original development of the CVGM. Second, the degree of certainty of fracture predictions is quantified. In combination, these features significantly enhance the robustness of the framework within which the model is implemented. Although this paper applies this approach in the specific context of the CVGM, the method can be generalized to other models that share similar characteristics.