Structural vibration-based identification of delamination in CFRP cylinders using complex frequency domain correlation and CNN

Abstract

A novel method based on complex frequency domain correlation (CFDC) and convolutional neural network (CNN) is proposed to identify the delamination in CFRP cylinders. Firstly, the CFDC based on the fusion of FRFs at multiple measuring positions of cylinders is constructed to characterize the delamination. Subsequently, the normalized CFDC is used as the input of a CNN model, so as to predict the circumferential angle and axial distance of delamination in cylinders. Moreover, the linear model between the complex frequency domain correlation indicator (CFDCI) and the delamination area is established due to the high linearity between them. Finally, the area of delamination is predicted based on the linear model, whose coefficients are predicted by another CNN model. The proposed method performs well in localization and area evaluation of delamination with high accuracy in the numerical cases. The delamination in each experimental case is identified within certain error even with serious noise in measurement. In addition, the comparison with other common methods indicates higher accuracy of the proposed method in delamination identification for CFRP cylinders.