Experimental studies of guided wave damage identification in beams using a probabilistic approach

Abstract

This paper presents an experimental study of quantitative identification of damages in beams using guided wave. A Bayesian probabilistic approach is proposed to identify the damage location, length and depth. The damage identification is achieved by solving an optimization problem, in which the probability density function (PDF) of the damage parameters is maximized. In this study a hybrid particle swarm optimization (HPSO) algorithm is employed to guarantee the global optimum solution . One advantage of the proposed methodology is that the Bayesian approach not only pinpoints the location, length and depth of the damage, but also quantifies the uncertainties associated with the damage identification results through calculating the posterior PDF of the identified damage parameters. This provides essential information for making decisions on necessary remedial work. In the experimental study a piezoceramic transducer is used for excitation. The guided wave signals are then measured using a laser Doppler vibrometer system. Metallic beams with different damage configurations are considered in the experimental verification.