Ensuring the Reliability of Damage Detection in Composites by Fusion of Differently Classified Acoustic Emission Measurements

The overall system reliability of complex or safety critical systems is of increasing importance. Composite materials are increasingly used instead of conventional construction materials due to advantageous mechanical properties, such as high specific strength resulting from their complex structure. Therefore, different nondestructive testing techniques, which indicate developing damages in composites, are proposed for continuous monitoring to ensure reliability and safety. Here passive emitted Acoustic Emission (AE) in combination with frequency analysis is used to identify characteristic features related to different failure modes of composites. The use of different filtering and classification methods to identify damages in composites results in various performance of classifiers. Ensuring high accuracy of the decisions evaluating faults, system changes or even the State-of-Health, the assignments from different classifiers evaluating the same situation can be fused to one final decision stating a similar statement but with improved probability of belief. Some factors just influencing only one individual filtering or classification can be compensated. Using ensemble selection strategies, only suitable classifiers are selected to be considered in the fusion process. In this contribution specific selection strategies and fusion algorithms are used to combine results of differently classified Acoustic Emission measurements to evaluate and distinguish different kind of damage of composite structures. The individual properties of suitable classifiers to be fused as well as the kind of selection are crucial for the overall accuracy. The results show the advantages of the combination of different assignments to ensure the reliability of a monitored system as well as the SHM system for monitoring.