Review of models and simulators for NDT reliability (POD).

With the use of probabilistic and risk-based methods by industry it is often necessary to estimate and input values for the reliability of inspection methods typically in terms of probability of detection (POD). In many cases this has been based on expert judgement. In recent years a more structured approach has developed using multi-parameter POD models. POD models have been around since the early 1990s. Inspection reliability has featured recently in a number of important legal cases, which has focused attention on the factors affecting reliability and approaches used to estimate reliability in practical circumstances. This paper will review the development of models for POD in Europe and the USA in the last decade and the current state-of-the-art in inspection reliability modelling by ESR Technology and others. Examples will include models for composite inspection developed for the European Space Agency and a specific model-based and simulator ‘POD Generator’ approach developed in the offshore industry through the HOIS joint industry project. This allows the POD for corrosion mapping and radiography inspections to be estimated and parametric studies undertaken to estimate the reliability of commercial inspection equipment in real situations. Other model-based approaches will be assessed. These include model-assisted POD approaches developed in the US aerospace industry, use of transfer functions to extend hit/miss and â versus a POD trial data to new applications, simulated POD trials, the addition of POD modules to established 3D and CAD inspection models and the use of NDT simulators. There are analogies now for POD modelling for NDT methods to the use of finite element analysis for stress analysis. In the early 1990s, FE analysis was treated with some suspicion and normal practice was to carry out expensive large-scale component tests. Nowadays, FE analysis is widely accepted and large-scale testing is carried out as the exception to validate the FE models. The use of model-assisted and model-derived methods to derive POD data has gained significant acceptance and arguably is becoming the preferred method to generate POD values, with POD trials providing model validation. Empirical methods developed in NNDTC to correct POD data for human and environmental effects are also discussed. [ABSTRACT FROM AUTHOR]