Vibration-based Damage Detection and Health Monitoring of Bridges.

摘要

This dissertation presents the findings of a research program that was conducted to develop a vibration-based technique that can be used for identifying damage locations in steel bridge girders. Most of the vibration-based damage detection methods that are available for identifying damage locations in bridges are physics-based techniques which are either based on simple modal characteristics, modally-based derived indices, or Finite Element Model updating. Although these methods have shown promise, problems such as low sensitivity of damage indices to local damages, high statistical uncertainty of the derived parameters, and model uncertainty for complex structures are major hindrances in developing practical plans for health monitoring of bridges. These problems are even augmented when the only practical source of vibration is ambient vibration.;This research program investigates two non physics-based approaches for potential vibrationbased damage detection in steel bridge girders under ambient vibrations. A novel damage diagnosis method was developed after modifying another damage diagnosis method published in the literature. Both methods are presented in the context of a statistical pattern recognition problem. Damage features were extracted in the 1st damage diagnosis method by fitting autoregressive and autoregressive with exogenous inputs models to the vibration responses measured at individual sensor locations. In the 2nd damage diagnosis method which is proposed as the result of this study, multivariate autoregressive models are used to extract the damage features from the vibration responses measured at multiple sensor locations at different damage conditions of the beam. Appropriate statistical measures have been proposed to capture damage-related deviations in the extracted damage features in both damage diagnosis methods. The damage locations were determined by finding the sensor location which corresponds to the largest deviations in the response. A critical threshold is also used in identifying the damage locations to prevent false-positive damage identifications.;The concept is demonstrated using a simply supported two span steel beam in the lab. The beam was instrumented with accelerometers and subjected to ambient vibrations. The ambient vibrations were simulated by applying random loads on the beam using a hydraulic actuator. Different sets of randomly generated numbers were used as inputs to simulate the ambient vibrations. Damages were induced in the beam by cutting different sizes of the bottom flange at two different locations. Also, the effect of varying temperature was investigated on the vibration responses. The recorded vibration time histories under different loading and structural conditions were used along with the proposed damage diagnosis techniques to identify the damage locations. The research findings demonstrate that the 1st damage diagnosis method is able to locate the general damage region while the 2nd proposed method successfully identified the exact sensor located closest to the physical damage locations.;Also presented in this dissertation is the result of a field study to investigate the extent and possible reasons for the daily modal variability which is observed in a two span simply supported steel-concrete bridge located in Lumberton, North Carolina. Less attention has been paid in the literature to the possible reasons for the observed variations in the modal properties of bridges induced by temperature variations. Vibration, temperature and displacement measurements were performed in this study during a 24 h period. Linear and nonlinear finite element analyses were performed to find the possible reasons for the observed variation in the natural frequencies. The findings show the different displacement contours of the bridge deck induced by temperature gradients in the cross-section of the bridge during night and noon time can contribute to changes in the stiffness of the bridge deck and variations in the measured natural frequencies.