Extraction of Full-Scale Dynamic Properties from Short Duration Records

Globally, urban environments are rapidly developing and high-rise structures are becoming more prominent. The design of these structures is often governed by wind-induced motions where habitability concerns, rather than strength, govern. As such, their responses are inherently sensitive to the dynamic properties of the system, particularly damping. Despite the many advances in finite element modeling, these parameters, and especially damping, are still not known exactly until after completion of the building, at times yielding structures that fail to meet habitability limit states. In efforts to provide a more reliable design-stage damping estimate, researchers began assembling databases of full-scale damping observations in hopes of developing predictive tools or at minimum guidelines to better inform designers. In addition, while finite element models help to provide a reasonable basis for natural frequency estimates in the design stage, various modeling assumptions do bear significantly on this process and would also benefit from in-situ validation. To conduct these validations, full-scale observations need to be collected from a wide range of structures with varying lateral systems. This thesis contributes to that effort by extracting dynamic properties from a 67-building Korean full-scale database. While this database encompasses a number of structural systems and materials, the buildings were only instrumented for short periods of time under ambient vibration levels, making the extraction of parameters like damping quite difficult. This thesis explores the performance of popular system identification tools like half power bandwidth estimates from power spectra and analytic signal components from random decrement signatures on these short duration records. To particularly address the competing demands of bias and variance minimization in power spectra, the thesis also introduces a maximum likelihood estimator with more intuitive frequency limits. Armed with this suite of analysis tools, dynamic properties are extracted from the 67 buildings in the Korean database, comparing performance across the methods and with previous system identification efforts on this same collection of buildings. This endeavor particularly underscored the challenges of analyzing short duration records and shortcomings of point estimates of damping from high variance spectra and random decrement signatures. Once the reliability of the extracted dynamic properties was established, this thesis explored the influence of characteristics such as height, aspect ratio and structural system on the dynamic properties, demonstrating how damping and frequency characteristics for structural systems dominated by cantilever effects are more appropriately parameterized by variables such as height or slenderness, while those from systems dominated by frame action are more reliant on floor plate aspect ratio. As more high quality full-scale observations become available, the analysis framework presented in this thesis and the new insights into parameterization of extracted dynamic properties can be applied to ultimately develop more reliable predictive models for use in design.