An autonomous strain-based structural monitoring framework for life-cycle analysis of a novel structure

Abstract

In recent years, developed societies have largely adopted smart systems operating on the basis of information extracted from data. For infrastructure systems as well, structural health monitoring (SHM) has long advocated a data-driven scheme for facilitating the operation and maintenance of infrastructure. In materializing such a goal, this paper demonstrates the procedures and outcomes of a SHM framework employed on an unconventional structure, namely, the recently built “Kaeng Krachan” Elephant Shelter at the Zurich Zoo, relying on a deployed set of Fiber Bragg Grating (FBG) strain sensors. The structure comprises an 80-m span free-form timber-composite cupola, carried by a post-tensioned reinforced concrete (RC) ring. FBG strain sensors are embedded into the ring in close vicinity to critical regions, selected in collaboration with the design engineers. The continuously acquired strain data are then exploited for extraction of performance indicators, relying on implementation of output-only identification methodologies. To this end, a non-parametric and a parametric output-only method, namely, a principal component analysis (PCA) scheme versus a Vector AutoRegressive (VAR) model, are employed and compared. Preconditioning of the predictive model is performed on the “healthy,” or undamaged, state of the structure, and the misfit between model predictions and subsequent measurements is exploited as a damage precursor. In this case, the VAR scheme proves a more robust representation of the measured strains, when compared against PCA, as a result of its inherent feature of memory.