Damage identification in beam-like structure using strain FRF-based damage index and artificial neural network

Abstract

A novel approach for structural damage identification of beam-like structures based on strain frequency response functions (SFRFs) using artificial neural networks (ANNs) is proposed. The dominated singular values of the phase space matrix derived from SFRFs are reserved as the features to denote the characteristics of the original SFRFs. And the reserved singular values mean change ratio (RSVMCR) is described as a new damage index for damage detection. The RSVMCRs and the elemental residual stiffness vector are utilized to establish the ANN model. Two numerical examples are investigated to verify the feasibility and effectiveness of the proposed approach.