As part of the on-going development of a new large-break (LB) loss-of-coolant-accident (LOCA) probabilistic pipe fracture mechanics code for the U.S.NRC, one of the most important aspects is the consideration of subcritical circumferential stress corrosion cracks (SCC), particularly the statistical characteristics of initiation of multiple cracks. As a result of multiple cracks initiating and growing, a longer surface crack could occur in a high residual stress field, such as at a girth weld, which could make a significant contribution to the probability of a LB-LOCA occurring. An important aspect for any probabilistic fracture mechanics code is to benchmark its deterministic results against real service data. This is first being done for intergranular stress corrosion cracks (IGSCCs) that occurred in BWRs in the past and then be extended to PWSCC cracks in PWRs. There were many cases of IGSCC cracks, and making sure that the general stress-corrosion-cracking model can predict those past cases will lead to the credibility of the model for future stress-corrosion cracking predictions. The chemical and mechanical process of SCC initiation is very difficult to define in a deterministic sense due to the complex and uncharacterized processes involved. Since it is generally assumed that SCC crack initiation times can be characterized by a statistical process, distributions can be developed to account for the uncertainty in the deterministic process. In this case, the Weibull distribution, which is considered to be a very flexible statistical model, is used in the current code to describe the initiation time distribution. In general, each initiation site can be sampled from the distribution to determine SCC crack initiation times. However, this model assumes an independency of multiple initiation sites, which may not be the reality since biased spacing between the cracks may exist circumferentially due to the pipe material, residual stress, applied loads, fabrication procedures, and environment aspects. In addition, Weibull model is not always the best choice to reflect the actual distribution while more fitting parameters are necessary for a complicated curve. This paper presents results of statistical analyses of length, location and spacing of circumferential IGSCC cracks in 43 girth welds from the Nine Mile Point Unit 1 plant’s main recirculation line that were removed from service in 1982 after 13 years of service. A measurement of the shortest circumferential distance encompassing all the cracks for a girth weld was proposed based on the intuitive judgment that if the cracks are clustered, the distance is shorter than otherwise if the cracks are randomly located. The analysis conducted shows the actual crack locations are slightly biased. Furthermore, the initiation time was back calculated for each crack length based on the typical crack growth model with a best-estimated residual stress field. The analysis of all the initiation times shows that a time distribution by a two-segment curve fitting can reproduce distribution of the crack length precisely, while the pure Weibull fitting failed to give such a result.

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