Statistical estimation of flaw size measurement errors for steam generator inspection tools

Accurate sizing of flaws in steam generator tubes is a critical component of a maintenance and inspection program. Knowledge of the measurement error of an inspection tool is important for determining the flaw severity, assessing the tool vendors claimed accuracy, as a component of the tools development program, as a reference for other inspection tools or candidates, and for probability of detection assessments. Often, tool reporting of flaw sizes is compared with data obtained from a reference tool or from a destructive test. The reference tool or destructive test data are generally assumed free from measurement errors but in a true sense they may not be so. It is therefore an advantageous and useful exercise to determine individually the measurement errors of each measuring system used. Statistical procedures commonly used to assess the quality of inspection tools estimate the total scatter between any two instruments, neglecting repeated measurements or more than two tool situations. One possible advancement is the statistical decomposition of the total scatter given by the root mean square error/differential into measurement error components to be associated with each of the instruments used. Important recent developments, such as Bayesian estimators of measurement errors, are included in this expository article to familiarize the researchers working in the nuclear industry with the state-of-the-art. (author)