In-service highway bridge condition assessment using high-rate real-time wireless sensor networks

This thesis primarily addresses two key aspects within the greater field of real-time vibration-based structural health monitoring. In its most basic essence, a structural health monitoring or condition assessment system ultimately consists of three tiers of information acquisition and advanced processing: response measurement and data aggregation, damage diagnostics for condition assessment, and prognostication for the ultimate prediction of the effect on the future performance and useful life cycle of the structure. The first aspect addressed in this dissertation is the real-time acquisition of lossless measurement data at sufficiently high spatial density and sampling rate to enable structural health monitoring with wireless sensor networks. The prototype system developed is the first known implementation of a lossless, high-rate wireless sensor network for highway bridges in a scalable architecture of up to forty sensors channels per star topology network. By reducing the cost and duration of testing as well as increasing the capabilities of the state-of-the-art wireless sensing technology, the platform has served not only to enable advanced academic research into the behavior and response of highway bridge designs but also advances a case for the eventual widespread adoption of advanced, sensor-based technologies for periodic of continuous quantitative assessment of in-service structures. The measurement accuracy and robust performance of the network transmission protocol are verified through laboratory experiments as well as field deployment on a multitude of different bridge designs. The field testing encompasses four unique cases of ambient vibration monitoring utilizing output-only system identification techniques to extract modal parameters from experimental models developed through advanced subspace methods. Additionally, strain measurements consistent with experimental load testing for inventory rating were acquired to supplement the dynamic response measurements and accentuate the multi-sensor capabilities of the platform design. In addition to serving to illustrate the wireless sensing approach, assess the real-world data quality, and validate the network transmission protocol, the field measurements provide unique case studies of the in-service behavior of several common bridge designs that, particularly in the case of integral abutment and skew bridges, lack extensive complementary experimental studies.;The second aspect addressed in the thesis is the incorporation of characteristic structural health monitoring response measurements into a diagnostic routine for in-service assessment of structural deterioration and early warning of imminent failure. Towards these aims, an experimental study was conducted on an end-of-service life highway bridge span that introduced progressive, prescribed damage to several elements including a bearing and multiple diaphragm connections. Stochastic subspace identification is used to develop an experimental model based on the underlying governing equation for dynamic mechanical systems from which modal parameters and contributions are evaluated. The state-space model is transformed to a steady state Kalman filter representation for one-step forward prediction of the response of future time histories. Prediction errors for various damage scenarios are then assessed as an indicator to damage through statistical analysis of the probability density function and comparison to the baseline model. The conclusion from the study is that the outlined approach displays remarkable ability to identify the onset of damage and localize the source, while providing strong evidence of the potential for indication of damage severity.