PoD curves, their derivation, applications and limitations

There is a large amount of Probability of Detection (PoD) data available (for example National NDT Centre (UK), NORTEST (Norway), NIL (Netherlands) and NTIAC (USA)). However, it is believed that PoD curves produced from PoD data are not very well understood by many who use and apply them. For example, in producing PoD curves, a certain material and thickness may have been used and yet one can find the same PoD quoted for a range of thicknesses. In other cases, PoD curves may have been developed for pipes, but they have been applied to plates or other geometries. PoD data are also highly dependent on the Non-Destructive Testing (NDT) methods used to produce them and these data can be significantly different, even when applied to the same flaws and flaw specimens. It is often assumed that the smallest flaw detected is a good measure of PoD, but there is usually a large gap between the smallest flaw detected and the largest flaw missed. Similarly, it is often assumed that human reliability is a very important factor in NDT procedures, and yet it is usually found not to be as important as other operational and physical parameters.

It is important to question the validity of how PoD curves are derived, applied as well as their limitations. This paper aims to answer such questions.