A Novel Hybrid Technique to Identify the Minor Cracks in the Structure During Online Health Monitoring

Abstract

The basic hypothesis of a structural health monitoring (SHM) system is that the global parameters like stiffness, mass, and damping of a structure are modified by damage and so the dynamic response. Nevertheless, in the case of subtle incipient damage, the changes in the dynamic characteristics of the structures developed in the structure alter only a few modal responses that too in a very mild manner, while all other modal responses remain intact. The damage features present in the modal responses of those few modes will be hidden in the measured total raw dynamic signatures. Often they get buried in the measurement noise. Keeping this in view, in this paper, a hybrid damage diagnostic algorithm combining a multivariate analysis technique called blind source separation (BSS) with time series analysis to identify and locate the minor cracks in structures is proposed. BSS decomposes the measured acceleration time-history responses into modal responses. We use an automated algorithm to isolate the modal responses which are sensitive to the presence of minor/subtle damages. These isolated modal responses are then reconstructed using the mixing matrix to obtain a new time-history data which is enriched with the minor damage features. The presence and spatial location of damage are obtained by processing the reconstructed time-history data using autoregressive–autoregressive with exogenous input (AR-ARX) model, using a Density Function of Probability (PDF), of the prediction errors. Both the numerical simulation studies and experimental studies are carried out to test and evaluate the proposed damage diagnostic algorithm and their capability in identifying minor/incipient damage like subtle cracks under noisy measurements.