Probabilistic Prediction of Crack Depth Distributions Observed in Structures Subjected to Thermal Fatigue

The application of risk-based technologies not only to in-service inspections but also to the design of components and systems, encompassing a plant life-cycle, is the way to be pursued for the improvement of design of new reactors such as fast breeder reactors. When doing so, it is necessary to develop an analytical method that is capable of estimating failure probabilities without a failure database that can only be established on the long-time accumulation of operational experiences. The prediction method should estimate failure probabilities based on actual mechanisms that cause failure. For this purpose, this study developed a structural reliability evaluation method using probabilistic prediction of crack depth distributions for thermal fatigue, which is one of representative failure modes to be prevented in components of nuclear plants. This method is an extension of probabilistic fracture mechanics approach but is capable of modeling crack initiation, crack propagation, and crack depth density distribution at a given cycle. To verify the methodology, crack depth distribution observed in thermal fatigue test specimens was evaluated, and it was shown that the method could reproduce the observed crack depth distributions fairly well. This is considered to explore the possibility that probabilistic fracture mechanics approach can be verified by experiments, which was deemed impossible so far. Further improvement such as explicit implementation of interaction mechanisms between adjacent cracks will allow this methodology to be applied to the procedure of optimization of in-service inspection planning, as well as to the optimization of safety factors in component design of nuclear plants.