Acoustic emission detection of filament wound CFRP composite structure damage based on Mel spectrogram and deep learning

This paper investigates the hoop tensile mechanical properties, damage mode classification, and damage evolution process of the filament wound CFRP composite structure with different winding angles by acoustic emission and deep learning. Firstly, a clear correspondence between damage modes and peak frequency ranges is established by using k-means clustering results and eight frequency band ranges obtained by wavelet packet transform. On this basis, three damage modes are characterized according to the energy content of different frequency bands. Subsequently, the trained Mel spectrum-CNN classification model is used to identify damage modes and analyze the damage evolution process. The results show that the hoop tensile strength of the filament wound CFRP composite structure with ±55° is higher than that with ±35°. The primary damage modes during the tensile process are matrix cracking and fiber/matrix debonding. The Mel spectrogram-based damage identification method can effectively classify damage modes. In addition, the combination of acoustic emission and deep learning shows significant potential for damage identification and monitoring.