Probabilistic Damage Characterisation in Beams using Guided Waves

This paper presents a model-based probabilistic damage characterisation methodology for beams using guided waves. Damage location, length, depth and Young's modulus of the material are treats as unknown parameters and are determined using optimisation approach to maximise the probability density function of the damage scenario conditional on the measured guided wave signals. A two-stage optimisation strategy is proposed using simulated annealing to first guarantee the solution is close to the global optimum, a standard simplex search method is then employed to precisely determine the global optimum. In addition to damage characterisation, the proposed methodology is developed based on the Bayesian statistical framework, hence, it allows quantification of the uncertainties associated with damage characterisation results. The accuracy and robustness of the proposed methodology are investigated through a series of numerical case studies in which the spectral element method is employed to model the wave propagation and scattering at the damage.