Vibration-Based Damage Detection For Timber Bridges

Vibration-based damage detection (VBDD) methods are global response-based methods that have the potential to provide valuable insight into the health of a structure. Dynamic characteristics obtained through vibration testing, such as the natural frequencies and their associated mode shapes, are directly related to both the stiffness and the mass of the structure, which are both good indicators of damage. In a typical application of VBDD methods, the vibration characteristics are obtained periodically to detect small changes in the response, indicating damage over time. However, this thesis considers using a snapshot of the vibration signature based on a single set of measurements to detect specific types of damage by comparing the response with that of other similar structures with known condition states. Like many provinces, Saskatchewan currently has a large inventory of aging timber bridges that are at or nearing the end of their service life. Many of these bridges are experiencing decay of their substructure elements (piles, pile caps and abutments), yet these are not always accessible for the inspector to identify. Furthermore, current inspection methods require lengthy and thorough site visits to reliably assess the condition of the timber bridge. Given the length of the current inspection methods and the large inventory of timber bridges in the Saskatchewan road network, other assessment tools are being sought. The objective of this thesis was to examine the feasibility of using vibration-based methods to assess the structural integrity of short-to-medium span timber bridges. Specifically, this thesis investigates the influence of realistic substructure stiffness on the dynamic properties of a timber bridge. Further research was conducted to determine if substructure deterioration could be detected reliably using the response from a single vibration test without the benefit of baseline (prior to damage condition) data. Additional variables, such as superstructure damage, superimposed mass on the timber bridge (to simulate the wearing surface), and interactions between the substructure/superstructure, were considered in this thesis. Furthermore, practical applications were studied, which included using limited sensors and impact excitation, as well as a study that used pattern recognition techniques in conjunction with a database of vibration signatures from various substructure condition states to assess the health of a timber bridge’s substructure. It was concluded that the first flexural mode shape could be described by deconstructing the mode shape into superstructure and substructure components. Based on the relative amplitudes of these components, differential and uniform support movements were used to describe the stiffness of the substructure. Additionally, a limited pattern recognition study, using neural networks, classified the integrity of timber bridge substructures on the basis of a single measurement of the bridge’s vibration signature. Variables such as superstructure damage, superimposed mass, and excitation type had relatively little influence on the results reported in this thesis. Furthermore, it was found that substructure and superstructure damage could be detected independently; however, superstructure damage detection required a baseline response