Abstract：This paper presents a layered damage detection method for carbon fiber reinforced plastic composite (CFRP) based on deep learning object detection algorithm. According to the hierarchical damage of CFRP under high frequency excitation, the local steady-state wave field will change. The object detection function of deep learning convolutional neural network is used to identify and locate the damage. In order to improve the generation efficiency of damage mode training samples, according to the local response characteristics of layered damage, this study adopted the transfer learning scheme and selected the isotropic aluminum plate structure with blind hole damage as the numerical model to replace the composite layered structure, which greatly improved the efficiency of numerical calculation. First, the YOLOv5s network is trained with the problem wave field diagram and corresponding label in the aluminum plate with damage. In the process of damage detection, the neural network will be trained by the input of the damaged CFRP steady state wave field excited by piezoelectric sheet. The recognition results of steady wave field show that the steady wave field recognition algorithm based on deep learning target detection proposed in this paper can quickly and accurately identify multiple spall damage in different positions in CFRP after migration learning training, providing a new detection method for rapid and intelligent composite damage detection.
2023, Vol. 44 
