Limitations in the Structural Identification of Long-Span Bridges

Writers discuss an in-depth analysis of the limits in St -Id of large structures, including the mitigation of human errors and the control of uncertainty due to the complexity of large structures and their loading environment. After globally and locally calibrating a 25,218 DOF 3D FE model of a major bridge, the average error between the model and experimental frequencies is determined to be 2.3 % for the frequencies of the first 15 modes. The confidence in the simulated local characteristics such as the stresses and distortions is judged to be around 50-75%, based on local calibration results. It is important to note that inherent assumption in ambient monitoring data analysis is that the structure remains stationary, observable and reasonably linear during the course of the monitoring. The data collected during a large number of intermittent time windows over several weeks included periods of different levels of traffic-induced vibration, various wind environments and temperature ranges. Averaging such data represents an effective composite of various inputs and the corresponding structural states and behaviors throughout the course of the monitoring, therefore in effect smearing all non-linearity and non-stationarity and effectively linearizing the structure.