Waveform based clustering and classification of AE transients in composite laminates using principal component analysis

Acoustic emission (AE) transients in composite laminate tensile test specimens are analyzed by principal component analysis (PCA). Signals resulting from matrix cracking and local delaminations were acquired by broadband transducers. The aim was to study the clustering and classification ability of PCA based on the time history of the recorded AE events. An unsupervised clustering analysis showed that AE signatures from matrix cracking and local delaminations were separated in different clusters. Some differentiations within these two types of signals were also observed, From the classification based on a training set, a majority of the signals in the test set were given the correct class membership, The applied pattern recognition method is an objective method using the complete information in the AE transients and the investigation shows that it may be a useful complement in the field of non-destructive evaluation. (C) 2002 Elsevier Science Ltd. All rights reserved.