Derivation and use of probability of detection curves in the nuclear industry

The use of probability of detection curves to quantify NDT reliability is common in the aeronautical industry, but relatively less so in the nuclear industry, at least in European countries. The main reason for this lies in the very nature of the components being inspected. Sample sizes of inspected flaws tend to be much lower and it is often very difficult to procure or manufacture representative flaws in test-pieces in a high enough number to draw statistical conclusions on the reliability of the NDT system being investigated. Similar considerations led to the development of the ENIQ inspection qualification methodology, based on the idea of the technical justification, ie a document assembling evidence and reasoning providing assurance that the NDT system is indeed capable of finding the flaws which it is designed to detect. The ENIQ methodology has become widely used in many European countries and is gaining appreciation outside Europe as well, but the assurance it provides is usually of a qualitative nature. The need to quantify the output of inspection qualification has become more and more important, especially as structural reliability modelling and quantitative risk-informed in-service inspection methodologies become more widely used. To credit the inspections in structural reliability evaluations, a measure of the NDT reliability is necessary. A probability of detection (POD) curve provides such metric.

The purpose of this paper is to briefly review the statistical models proposed to quantify NDT reliability, to highlight the potential problems that can arise if the main underlying assumptions and requirements are not verified and to clarify the confusion that can arise over the nature of the POD curve and associated confidence bounds.