Cluster analysis of acoustic emission signals for 2D and 3D woven carbon fiber/epoxy composites

Understanding the failure mechanisms in textile composites based on acoustic emission (AE) signals is a challenging task. In the present work, unsupervised cluster analysis is performed on the AE data registered during tensile tests on 2D and 3D woven carbon fiber/epoxy composites. The analysis is based on the k-means++ algorithm and principal component analysis. Peak amplitude and frequency features – peak frequency for 2D woven composites and frequency centroid for 3D woven composites – were found to be dominant in cluster analysis. Cluster bounds were identified for both reinforcement types. These bounds do not differ for both reinforcement types and can be used as a starting point for AE analysis of other carbon fiber/epoxy composites. The statistics of high frequency AE events are compared with the estimates obtained from a fiber bundle model based on Weibull fiber strength statistics. The number of AE events agrees well with the number of groups of carbon fibers that fail simultaneously. This finding may provide a new way to explain why the Weibull distribution predicts much more fiber breaks than measured by AE.status: publishe