Damage characterisation in composite materials under buckling test using acoustic emission waveform clustering technique

Acoustic Emission (AE) is extensively used to assess damage for structural health monitoring of loaded structures. Knowing the correlation between the detected AE signals and a particular failure mechanism is important in order to help the operator of the structure to make a correct decision about the structures integrity. The purpose of this investigation is to study the clustering and classification ability of unsupervised clustering based on the signal waveforms, in order to identify and characterise the most critical damage mechanisms in a buckling test on composite panel. The Hilbert Transformation and Principal Component Analysis (PCA) were utilised to extract the waveform appearance information for use as an input vector to the clustering process. The unsupervised clustering was performed using the k-means method. The results show that there is a good agreement between clustering classes and the damage revealed by the ultrasonic C-scan image. The reliable indication of different damage modes (i.e. delamination and matrix cracking) occurring in different regions of the panel allowed a validation of the use of unsupervised classification techniques based on signal appearance in order to categorize the AE signals