Abstract:
This thesis addresses the issue of determining the structural modal parameters (natural frequencies, damping ratios and mode shapes) using the freely available ambient vibration measurements. The analysis methods considered for this purpose are the recently developed stochastic subspace identification algorithms that prove to be useful for ambient data analysis. These algorithms are essentially based on the assumption that the unmeasured excitation is the realization of a stochastic process, and try to fit state space models to the experimental data in the time domain. There are two different implementations of subspace methods. One of them converts the measured vibration signals to output covariances which can be considered as some sort of free decay response, and hence can be used to feed the realization algorithms originally formulated to treat impulse responses. The other algorithm, on the other hand, circumvents the covariance estimation step, and tries to fit a state space model directly on the raw output measurements by means of some projection techniques. The first application in this study deals with the comparison of the two different identification algorithms with regards to their modal parameter identification capabilities. The comparison is handled by designing a Monte Carlo experiment based on the data simulated from a simple spring dashpot model. The relatively better performing algorithm is, then, used for the modal parameter identification of the IASC-ASCE structural health monitoring benchmark problem. The problem contains both an analytical and an experimental phase, and some implementation issues are discussed herein. The final case study concerns the modal analysis of the Vincent Thomas Suspension Bridge via the acceleration signals obtained under the operating conditions of the bridge.
