STRUCTURAL ANOMALY DETECTION BASED ON PROBABILISTIC METRIC DISTANCE OF TRANSMISSIBILITY FUNCTIONS

摘要

Transmissibility function (TF) has been extensively used as damage-sensitive features in structural condition assessment. Based on the theoretical findings of circularly-symmetric complex Gaussian ratio distribution for transmissibility, the study proposes a new data-driven damage detection algorithm by accommodating multiple uncertainties of frequency responses. Based on the analytical probability density function of TFs of the healthy and of different possibly damage scenarios, a probabilistic metric is calculated as a damage index to identify the dissimilarity between the probability distributions of TFs under different states, which allows the automatic identification of structural anomaly. Numerical studies are carried out to verify the effectiveness and accuracy of the proposed methodology.