Lamb wave-based damage imaging of CFRP composite structures using autoencoder and delay-and-sum

Ultrasonic guided wave is a promising technique for structural health monitoring and nondestructive testing. However, due to the anisotropy and complexity of composite materials, the imaging performance of numerous signal processing methods deteriorates with significant artifacts and unsatisfactory accuracy. To obtain a better damage imaging performance of Lamb waves in noisy and noise-free conditions, a weighted delay-and-sum (DAS) imaging method based on denoising autoencoder (DAE) learning is developed for complex composite structures. The traditional DAS formulation is modified to be more compatible with anisotropic materials. The DAE with feature learning capability is then employed to extract potentially efficient features and remove noise from ultrasonic signals. Several verification experiments conducted on flat or curved and stiffened composite structures have confirmed the ability of the DAE-DAS method to suppress artificial noise and to intensify sin-gularities induced by the anomalies. By comparing with the unweighted DAS methods and the weighted DAS method without feature extraction, the proposed algorithm has satisfactory robustness to achieve higher local-ization accuracy and fewer artifacts.