Cross-scale data-based damage identification of CFRP laminates using acoustic emission and deep learning

Abstract

Damage characterization of laminated composites has undergone extensive research, leading to the development of several damage models that incorporate acoustic emission (AE). However, feature extraction from AE data will cause the loss of certain information in the construction of damage models, which manifestly falls short of feature analysis in the frequency domain. In this study, a novel cross-scale data-based damage identification methodology for carbon fiber-reinforced polymer (CFRP) laminates is proposed by combining the AE technique with deep learning approach. Elementary experiments involved with single damage mode are designed to avoid blind inference, and the cross-scale correlation of AE features between component materials and CFRP laminates is established by wavelet packet transform (WPT). The time–frequency spectrums of AE signals of CFRP laminates by continuous wavelet transform (CWT), which adequately preserve the frequency domain information, are adopted as inputs to the convolutional neural network (CNN) model. The proposed methodology achieves a high accuracy of 96.3% in detecting and classifying damage modes (i.e., matrix cracking, matrix/fiber debonding, and fiber breakage) in unidirectional CFRP laminates.