Guided Wave-Based Inverse Identification Technique for Damages in CF/EP Composites Using Neural Network Algorithm

摘要

A structural health monitoring approach, combining the model-based numerical simulation with non-model-based information processing techniques, was developed for the identification of impact-induced damages in CF/EP composites. Guided Lamb wave propagation in the quasi-isotropic laminates under various damage situations was simulated and corresponding spectrographic characteristics were purified, extracted and further compressed over the time-frequency domain via the wavelet transform, referred to as "digital damage fingerprint," which were then utilized for the offline development of an artificial, neural-network-based damage parameters database supervised by an error-back propagation (BP) neural algorithm.