Thermal aging of cast austenitic stainless steel (CASS) piping is a concern for long-term operation of nuclear power plants. Traditional conservative deterministic fracture mechanics analyses lead to tolerable crack sizes well below the sizes that are readily detectable in these large-grained materials. This is largely due to the conservative treatment of the scatter in material properties and the imposition of multipliers (structural factors) on the applied loads. In order to account for the scatter in the tensile and fracture toughness properties that enter into the analysis, a probabilistic approach is taken. Application of the probabilistic fracture mechanics (PFM) model to representative problems has led to questions regarding the dominant random variables and the influence of the tails of their distributions on computed failure probability. The purpose of this paper is to report the results of a study to identify the important random variables in the PFM model and to investigate the influence of the distribution type on the computed failure probability. Application of the PFM model to a representative piping problem to compute the depth of a part-through part-circumferential crack that will fail with a defined probability (10−6 for example) revealed that the fracture toughness was not a dominant variable and the distribution of the toughness did not strongly affect the results. In contrast to this, the flow strength (which enters into the calculation of the applied crack driving force — J) was important in that low flow strength was controlling the low probability failures in the Monte Carlo simulation. Hence, the low-end tail of the flow strength distribution was influential. Various types of distribution of flow strength consistent with the available data were considered. It was found that the distribution type has a marked, but not overwhelming, effect on the crack depth that would fail with a given probability. From this it is concluded that the PFM model is quite robust, in that it is not highly sensitive to uncertainties in the dominant input distributions.

This content is only available via PDF.
You do not currently have access to this content.