Damage Identification in Composite Structure Using Machine Learning Techniques Based on Acoustic Emission Waveforms

The demands for glass fibre-reinforced polymer (GFRP) composites are commonly perceived in lightweight applications because of their high specific strength and stiffness. When those structures are subjected to mechanical loading which lead to crack initiation and failure of structures. The most common damages in GFRP composites are matrix cracking (MC), fibre-matrix debonding (FMD), delamination (DL), and fibre breakage (FB). To predict those damages in the GFRP composite structure, acoustic emission (AE) signal features such as amplitude and time are taken from research that was already done by Choudany et al. Furthermore, using unsupervised machine learning (ML) algorithms such as k-means++ clustering and agglomerative hierarchical clustering (AHC), the obtained AE features are classified into different clusters. Those algorithms grouped different composite damages based on characteristics of AE waveforms in different clusters. A supervised ML technique k-Nearest Neighbour (k-NN) is used to predict the accuracy and reliability of k-means++ and AHC algorithms based on obtained confusion matrix.