An Experimentally Accessible Probability of Detection Model for Acoustic Emission Measurements

In testing of structural components by the acoustic emission method, the requirement arises for quantitative prediction of the probability of detection (PoD) of an acoustic emission signal. Motivated similar as for other nondestructive testing methods, the suitability of given experimental settings to reach a certain likelihood of not missing relevant signals should be predicted. In contrast to other nondestructive testing methods, two of the key factors are not only the equipment and the inspector, but also the variability of the acoustic emission sources and the attenuation effects. As the strength of crack-based acoustic emission sources cannot be changed arbitrarily in the experiment, their characteristic amplitude distribution is accounted for by generation of reference datasets in small laboratory scale specimens. This assumes datasets with 100% PoD for those signals at a particular propagation distance. The prediction of the resulting PoD at another distance in a structure is achieved by means of amplitude reduction based on the measured attenuation values. For the latter, approaches using constant attenuation factors and attenuation mapping approaches are evaluated and compared to an experimental assessment of the PoD values using artificial test sources. Based on the agreement of calculated and measured PoD values, the presented approach appears promising to predict PoD values in geometrically and acoustically complex structures.