Optimal Artificial Boundary Condition Configurations for Sensitivity-Based Model Updating and Damage Detection

Frequently, in structural system identification (model updating or damage detection), the available set of data is incomplete, both spatially and in modal content. This incompleteness leads to the solution of underdetermined linear systems. In order to improve the identification, additional independent measured data must be found. In the past, it has been shown that such data can be easily obtained from the application of Artificial Boundary Conditions (ABC), imposed on both the baseline FE models and the measured frequency response data. This can be accomplished without any physical modifications to the experiment and, hence, no additional expense on different systems, or more than once, in order to get the modal data needed for the analysis. In this thesis, the procedure of sensitivity-based structural system identification, using ABCs, and enhanced by parameter grouping/clustering, is examined. It is shown that the optimal sensitivity matrix is a square and diagonal dominant one, which can be used with quite accurate results both for localization of parameter errors, and the determination of the magnitude of parameter error. The numerous ABC configurations available, even from a small measured data set, allow an optimal sensitivity matrix to be found for many parameters. These concepts are demonstrated using simulated measurements along with finite element models.