Fuzzy clustering of stability diagrams for vibration-based structural health monitoring

A primary challenge to implementing structural health monitoring techniques on civil infrastructure is the identification of structural changes in the presence of natural changes in structural response due to environmental variables such as temperature. Data from the Z24 Bridge recorded over the course of nearly a year are analyzed in this article. Covariance-driven Stochastic Subspace Identification is applied to the data to identify the modal parameters of the structure. A large number of numerical poles are identified with the real physical poles. A Fuzzy Clustering Algorithm is then used to extract parameters indicative of the bridge's state from the mixture of real and numerical poles. The main benefit of this approach is the lack of need for mode shape information and thus its applicability to structures monitored with spatially sparse sensor grids. The method is shown to provide very encouraging results in separating the response data from the Z24 Bridge in healthy and damaged states in varying environmental conditions. The method does not explicitly identify the changes due to environmental variables but it is found that the changes in the parameters identified due to damage are greater than those due to environmental variability and therefore may be flagged. The procedure is also applied successfully to a second data set obtained from monitoring a tall building over several years of its early life to identify gradual or sudden structural changes.