A method for detection of delamination depth position within composite laminates based on 2D continuous wavelet transform and CNN

This paper proposes a laminate mode shape curvature (MSC) analysis method combining 2D continuous wavelet transform (2D-CWT) and convolutional neural network (CNN) technologies to address the delamination damage detection in composite laminated plates. This method constructs a new network model based on the emerging CNN technology and achieves good results by detecting delamination damage through learning the MSC images processed by 2D-CWT. The train in this study is constructed by inserting randomly generated delamination with varying geometric sizes, depth positions, and geometric positions into a specified 16-layer carbon fiber-reinforced plastic finite-element model. In addition, the method of establishing the finite-element model has been verified by experiments, the error of the simulation frequency is less than 10%, and the mode shape is consistent. The results show that the proposed method can effectively detect the depth position of the delamination with a detection accuracy of 93.89% ± 2.74% using the comprehensive dataset. Compared with the Resnet-50 backbone network, the proposed network improves detection performance by 0.86%. This finding expands the range of tasks that can be accomplished by mode shape detection methods and has broad prospects for subsequent engineering applications.