Multi-Axial Fatigue Crack Growth Uncertainty Quantification and Risk Management

Abstract

The modeling and risk management of fatigue crack growth is a problem of critical importance in any mechanical system. The work presented in this study demonstrates two efficient methods for the modeling of non-planar fatigue crack growth and also the quantification of uncertainty in the model prediction. Discretization errors, both temporal and spatial, are considered for each model. The model parameters for crack growth models typically are obtained via experimental data. A methodology is presented here to calibrate those parameters when experimental errors are considered. The consideration of experimental errors leads to a more realistic approximation of the model parameters. A methodology for model calibration and validation using multiple sources of information is accomplished through a Bayesian network approach. Finally, natural variability, data uncertainty and model uncertainty are considered to provide an informed framework for risk management decision making with respect to inspection scheduling and fidelity.