Integration of the Kohonen's self-organising map and k-means algorithm for the segmentation of the AE data collected during tensile tests on cross-ply composites

The acoustic emission (AE) technique is a useful way for the investigation of local damage in materials. This study deals with the ability of a Kohonen's map to classify recorded AE signals collected during tensile tests on cross-ply glass/epoxy composites in order to monitor the chronology of the damaging process. An unsupervised clustering analysis shows that AE signals are distributed into three clusters. The proposed two-stage procedure is a combination of the Self-Organising Map (SOM) and the k-means methods. In the present work, Kohonen's map is applied as an unsupervised clustering method for the AE signals generated in cross-ply composite specimens during tensile tests. The input vectors of the signal descriptors used in the clustering procedure are calculated from the signal waveforms. The k-means method is then applied on the neurones of the map in order to delimit the clusters and to visualise the topology of the map.