Real Defects as the Major Challenge of Useful POD Evaluations

For critical components, the use of nondestructive testing (NDT) systems and their evaluation with probabilistic tools is already a standard procedure for deciding whether the system is capable to detect the critical defects. Often the only available data, which are usable are artificially made reference defects. The reasons for not creating a large amount of real critical defects are economical and also of technical nature: The statistical approaches require a large amount of data and at the same time a wide knowledge about the criticality of the defect and the physical behavior of the NDT system. There are major advantages to use artificially made reference defects: The manufacturability is guaranteed and, therefore, the defect parameters are known and the costs of the manufacturing process are calculable. The disadvantage is the validity of the defects: Only through a trustworthy use of technical justification and the opinion of experts these artificial defect data can create information for the later use in the reliability studies. This fact is often forgotten in the probabilistic evaluation of NDT systems. With the use of real and realistic defects the complete design of experiments and their evaluation of the process to get a probabilistic evaluation becomes more difficult. Not only the measurement of the defect size through the master NDT methods or metallographic methods increase the complexity of the evaluation, but also the relationship between the defect and the signal parameters get more sophisticated. This work will show that the introduction of real defects into the probability of detection (POD) evaluation can be a challenge, which must and can be overcome. The focus of this overview will be on how to plan and conduct experiments for the evaluation of NDT systems. This will be shown on an example of radiographic, ultrasonic and eddy current testing. M or e in fo a bo ut th is a rt ic le : ht tp :// w w w .n dt .n et /? id = 21 79 2