Laplacian score and K-means data clustering for damage characterization of adhesively bonded CFRP composites by means of acoustic emission technique

Abstract Carbon Fiber Reinforced Polymer (CFRP) composites, adhesively bonded in a Single Lap Shear (SLS) configuration, are tested in this study. The damage characteristics are analysed using Acoustic Emission (AE) data recorded during the test. Among the different features of the AE data, the most suitable features are selected using a methodology developed based on the Laplacian scores for feature selection. The selected features are then analysed using k-means clustering algorithm for establishing a relationship among them. Amplitude of the recorded AE signals and their Frequency Centroid (C-Freq) had a strong relationship, which is used for characterizing the damage modes. The cumulative counts of the clustered data of Amplitude vs C-Freq are used for characterizing the damage modes. A comprehensive characterization of the damage modes in each stage of loading and the final failure is made using this method.