Statistical analysis of measured growth rates of intergranular stress corrosion cracks

Intergranular Stress Corrosion Cracking (IGSCC) has been a serious problem within the nuclear power industry all over the world. Substantial efforts have been expended on research and experiments to enhance the understanding of the mechanisms behind this type of corrosion. An increased knowledge of such mechanisms is a condition precedent for preventing or at least reducing the problem. The crack growth rate has been the focus for analysis and estimation. As the data bases created in laboratory and other experiments have expanded there has been a growing pressure to use these data for the assessment of detected damage, even though the primary objective of these data has not always been to quantitatively determine the propagation rate of stress corrosion cracks. One exception here is the international project known as the 'SKI Round Robin', initiated by the Swedish Nuclear Power Inspectorate in order to be able to compare in a more controllable way various laboratories' measurements of IGSCC propagation rates. The aim of this study is to present methods for the statistical analysis of qualified data of crack propagation rates for varying material, environmental and stress conditions. A good knowledge of such connections as well as an understanding of the uncertainty in these connections can be used to evaluate the progress of already detected cracks and as basis for revision of existing regulation programmes. The Swedish Nuclear Power Inspectorate applies, at present, the principle of an 'upper bound curve' to describe the crack propagation rate as function of the stress conditions. In the long run another type of reasoning might appear as a result of a deeper statistical analysis of crack growth rates. To be able to conduct the statistical analysis for several parameter combinations a preliminary computer code has been developed based on Bayesian statistical methodology. In this report possible statistical models are discussed and exemplified and various ways to present and interpret the results of the analysis are shown