Frequency response function based finite element model updating

This dissertation presents the development and application of a frequency response function based finite element model updating method. The societal problem that this research hopes to address is the continually aging infrastructure of the United States, particularly bridges. The method presented combines the usefulness of commercially available finite element modeling programs with advanced optimization techniques to solve the inverse problem in a robust and elegantly simple form. The method is applied to a simulated bowstring truss, where all stiffness, mass, and damping parameters are simultaneously identified. Grouping and ungrouping proved to be a successful technique to aid the method in detecting, locating, and quantifying change in structural parameters. A benchmark structure is then used to show the usefulness of the method to accurately detect change in structural properties of a physical structure with measurements. This example shows the robustness of the method in the presence of real modeling and measurement errors. Model calibration using measurements from the intact structure is an essential part of the process to mitigate incorrect assumptions made during modeling. The final application of the method is finite element model calibration of a full scale bridge using measurements from a dynamic bridge test. Issues of observability and identifiability of structural parameters are fully explored prior to calibration of the finite element modeling. Creating a refined finite element model was essential to the success of the model calibration. The hope of the author is that the methods and techniques presented here can be used in bridge structural health monitoring programs in the future.