A Multi-Model Based Approach for the Detection of Subtle Structural Damage Considering Environmental Variability

Abstract

Minor structural damages like incipient cracks are difficult to detect as they alter the structural stiffness marginally. It is difficult to extract the features of minor damage, from the measured time history responses, which are usually contaminated by measurement noise. Also, the effect of environmental/operational variabilities often misleads the damage diagnostic process, especially for subtle damages. To tackle all the above challenges in detecting and locating the minor incipient damage, an automated multi-model based data-driven technique is proposed in this paper. It is based on the fact that a subtle damage alters only some structural modes, while the others remain unaltered. Hence, the proposal here is to decompose the measured time-history response into the modal components and then reconstruct the signal using only the modal components with the damage sensitive features. The reconstructed signal is used in damage diagnosis. As an improved version of the second-order blind identification, the blind source separation technique is proposed in this paper for signal decomposition and a crisp automated algorithm is presented for isolating the modal components with damage sensitive features. The autoregressive moving average with exogenous input model with the cepstral distance as the damage index is employed for localizing the damage. The proposed multi-model approach is completely automated. Numerical studies have been carried out to demonstrate the effectiveness of the proposed algorithm. Also, experimental studies have been conducted to ensure the practicality of the proposed technique.