Confidence Assessment in Fatigue Damage Prognosis

This paper presents a methodology for assessing the confidence and the predictive capability of prognosis models, using the application of fatigue crack growth analysis. Structures with complicated geometry and multi-axial variable amplitude loading conditions are considered. Several models – finite element model, crack growth model, retardation model, surrogate model, etc. – are efficiently connected through a Bayes network and the parameters of these models are calibrated after collecting inspection data. The results of the calibration are then used to develop a Bayesian confidence metric to assess the confidence of the models used in fatigue crack growth analysis. Three types of uncertainty are included in analysis: (1) natural variability in loading and material properties; (2) data uncertainty, due to crack detection uncertainty, measurement errors, and sparse data; (3) modeling uncertainty and errors in crack growth analysis, and finite element analysis. The proposed methodology is illustrated using a numerical example of surface cracking in a cylindrical structure.