Statistical Initiation and Crack Growth Models for Stress Corrosion Cracking

Statistical distributions of initiation times and crack growth rates are needed for probabilistic fracture mechanics models. Times to failure in laboratory tests on small specimens are about a factor of 1000 shorter than the times to failure of comparably sized “specimens” in the field would have to be in order to get realistic component failure rates. Thus while specimen tests are useful in identifying parametric dependencies, it is unlikely that they can be used directly to develop initiation models for field components without using field data. A scaling approach is proposed to provide a method for pooling data from different size components and for extrapolating experience from one set of components to another set. Estimates of statistical distributions for initiation of stress corrosion cracks are developed from field data for BWR pipe cracking and CRDM cracking. Estimates of statistical distributions of crack growth rates are developed by combining phenomenological models for crack growth rates with expert judgment on the range of input parameters to those models.