Bayesian model updating approach for experimental identification of damage in beams using guided waves

A Bayesian approach is proposed to quantitatively identify damages in beam-like structures using experimentally measured guided wave signals. The proposed methodology treats the damage location, length and depth as unknown parameters. Damage identification is achieved by solving an optimization problem, in which a hybrid particle swarm optimization algorithm is applied to maximize the probability density function of a damage scenario conditional on the measured guided wave signals. Signal envelopes extracted by the Hilbert transform are proposed to minimize the complexity of the optimization problem in order to enhance the robustness and computational efficiency of the damage identification. One of the advantages of the proposed methodology is that instead of only pinpointing the multivariate damage characteristics, the uncertainty associated with the damage identification results is also quantified. This outcome provides essential information for making decisions about the remedial work necessary to repair structural damage. The experimental data consist of guided wave signals measured at a single location of the beams. A number of experimental case studies considering damages of different scenarios are used to demonstrate the success of the proposed Bayesian approach in identifying the damages. The results show that the proposed approach is able to accurately identify damages, even when the extent of the damage is small.