Reliability of nondestructive inspection

Abstract

This report describes the concept of reliability of nondestructive inspection (NDI) and reviews the different aspects involved. First, the probability of detection (POD) of flaws in flawed specimens is discussed. To provide a measure of confidence in an estimated POD, lower botind values of this probability at a given confidence level are calculated with statistical methods. The inspection sample size in practice limits the applicability of statics. Therefore the available data points are grouped in intervals of flaw size; four interval methods are discussed in this report. Also a different approach for the representation of NDI reliability, by making use of linear regression analysis, is described. Next, the more general quadrinomial distribution of the possible outcomes of NDI, with successful and unsuccessful inspections of both flawed and unflawed specimens, is discussed. In analog with the POD for the flawed specimens, a probability of recognition (POR) is defined for the unflawed specimens. Both POD and POR must be included in a characterization of NDI reliability. Possible measures for this characterization are discussed. It is concluded that it is doubtful whether a single parameter will give an appropriate measure of NDI reliability; instead it is recommended to establish minimum values for both the POD and POR at a specified confidence level. Finally, some remarks are given about the application of reliability demonstration programs for the determination of flaw detection capabilities of various NDI-methods.