Human factors in POD modelling and use of trial data

Abstract

In recent years there has been increased use of POD modelling to complement, and as an alternative to, probability ofdetection (POD) trials. A number of such models have been developed in ESR Technology's National NDT Centre (NNDTC) in the UK, at Iowa State University in the USA and at other institutions. Such models allow better planning of POD trials, can generate POD estimates in their own right and enable the sensitivity to individual parameters to be assessed. They are also a valuable training and set-up aid. Different methods are discussed that have evolved to take account of human error and environmental factors (HF) in the use of POD data from trials or model estimates. These methods are discussed in the context of the multiparameter POD and human reliability model developed in the European-American reliability workshops. Progress on validation of these methods is discussed. These approaches are discussed and a number of specific examples are given. These include: (i) use of empirical human factor data such as from the PISC III or PANI III trials; (ii) use of models and simulators developed for ESA and HOIS to carry out simulated POD trials ( ‘spot the flaw’); (iii) comparison of POD data from automated and manual inspections; and (iv) an axle simulator model which allows real crack indications to be added to A-scans from axles. A number of examples will be given utilising probability of detection (POD) and NDT simulation models that ESR Technology originally developed for the European Space Agency. These have recently been developed and customised for the offshore oil and gas industry through the HOIS joint industry project and also for railway axle inspection. The models have covered ultrasonics, radiography and other methods using a modular approach that can take inputs for signal, background noise and detection criteria from models or experimental data, reproducing what is done in practice. A feature of the models is that they produce simulated data, so the theoretical estimates can be compared with POD estimates in simulated trials for decision and analysis using the 'spot the flaw' method, providing insight on human reliability issues related to interpretation of data. Input data can be included representative of commercially available NDT equipment. In some areas such as railway axle inspection or turbine blade inspection inspectors rarely see a defect, and defects can be missed because of the unfamiliarity with how these appear against background geometric echoes. Such models or simulations allow signals from real defects to be included, allow human factors to be considered and assist inspectors in understanding how real cracks may appear in inspection data compared with the reference reflectors that are often used in reference and calibration samples.