Comparison study of output-only subspace and frequency-domain methods for system identification of base excited civil engineering structures

Abstract

A number of practical issues must be considered when conducting system identification of civil engineering structures when using output-only data sets. In this study, system identification methods in the time- and frequency-domain are considered: stochastic subspace identification (SSI) and frequency domain decomposition (FDD), respectively. In order to evaluate these two output-only analyses, support-excited frame structures are selected as a common civil engineering structure of interest (e.g., building structures exposed to seismic loads). Due to the absence of a direct means of mathematically evaluating the results of the output-only analyses, this study proposes an indirect method of comparison. Specifically, an input-output system identification study is first conducted by the subspace system identification method to estimate a full set of state-space model system matrices. Then, time-domain comparisons between the predicted and measured responses are conducted. Next, conversion of the input-output model to modal space is executed to allow a direct comparison between modal parameters extracted from the two output-only models can be made. Detailed experimental testing using a steel frame structure exposed to seismic ground motion is conducted to allow the accuracy of the two output-only analyses to be evaluated.