Implementation of Supervised Classification Method of Acoustic Emission Signals: Damage Mechanisms Identification of Non-hybrid and Hybrid Flax Fibre Composites

In the present work, a classification approach of acoustic emission (AE) signals that combines unsupervised and supervised methods was presented to investigate damage mechanisms in hybrid fibres reinforced epoxy composites. These ends were elaborated by hybridising unidirectional (UD) flax and glass fibres and by using a vacuum infusion technique. First, specific tensile tests of various UD flax and glass fibres specimens were coupled with AE monitoring. This enables us to construct different datasets of unlabelled AE signals. Then, a clustering method, named incremental clustering (IC), was used to assign each AE signal with a typical damage mechanism according to its characteristic features. The decision rule of the IC was based on the principle of support vector data description (SVDD). The IC allowed the building of a training database that includes AE signals of five damage mechanisms, namely matrix cracking, flax fibre-matrix debonding, glass fibre matrix debonding, flax fibre breakage, and glass fibre breakage. Finally, the training database was coupled with the Support Vector Machines (SVM) classifier to identify the damage mechanisms of two hybrid flax-glass fibre composites. The experimental results indicate that the classifier model can predict the damage mechanisms with an accuracy of 93.13%. In conclusion, the proposed classification approach proved promising to investigate the damage mechanisms of hybrid composites.