The role of probabilistic approaches in NDT defect-detection, -classification, and sizing

Starting with the damage tolerance design of structures in aerospace applications it became obvious: NDT procedures, utilized according to standards in the regulated area and according to technical reports, guidelines and handbooks in non-regulated applications have to take into account the uncertainties of the techniques which are system features influenced by the type of product to be inspected, more over its individual design and manufacturing techniques, the type of non-conformity or irregularity to be detected, the type of ND technique selected, the hardware (transducer and system electronics), the software reliability and - the human factor of the inspector or user. All of the influences are summarized in a feature which is called POD - probability of detection. If there is a certain probability for detection, then there is also a PND, a probability for non-detection. As each technical system is not complete and not ideal the indication of a 'finding' or an 'alarm' by the system can be a real hit but also a false indication, or also called, false alarm. In the same way, in the case of a 'calm' system, i.e. no alarm is observed and no finding is indicated, the system' answer can be true or false. For each of the four cases the knowledge of the probability has to be known. POD distributions as function of the size of an individual material irregularity are relevant influencing parameters concerning the probability of the failure of a component evaluated according to a failure assessment diagram so far as a probabilistic approach is applied. The contribution introduces into the state of the art of a software development which especially in the future will be applied to evaluate pipeline integrity after performing in-line non-destructive inspections