Signature Recognition Of Acoustic Emission From FRP Structures


Proceedings - 7th ECNDT, Copenhagen, 26-29 May 1998, pp. 2295-2302

A. A. Anastasopoulos, S. J. Vahaviolos, J. C. Lenain


The paper demonstrates the use of Unsupervised Pattern Recognition techniques for the identification of the Signature of Acoustic Emission (AE) signals emitted from Composite Structures. The proposed method consists of procedures for descriptors selection, data clustering and techniques to validate the resulting partitions. The three previously mentioned essential steps for the application of the proposed pattern recognition technique in AE are presented and representative results from various composite structures, ranging from model specimens to FRP pressure vessels are discussed. The results prove the great help the method provides to the AE analyst in order to discover the statistical nature of both AE descriptors and signals, as well as to make useful correlation with the source of emission. The various stages of damage initiation, propagation and severity are successfully described by plotting the cumulative hits of the resulting classes vs. the applied load. Finally, the possibility of coupling the proposed method with existing data bases from field testing and/or other artificial intelligence techniques is discussed.