Experimental modal analysis of star-spar buoy using eigensystem realization algorithm and stochastic subspace identification methods

The star-spar buoy employed in this study is a small buoy system for omni-directional wave energy harvesting, and it has been designed for the system to resonate at a dominant period of 2.25 seconds in heave motion. However, this design frequency must be verified through experiments. The objective of this study is thus to obtain the frequency and damping, associated with the heave motion of the star-spar buoy through tank and sea experiments. Two specific time domain modal identification techniques to be utilized in this study are eigensystem realization algorithm (ERA) and stochastic subspace identification (SSI). ERA is a deterministic (input-output) modal identification technique and SSI is a stochastic (output-only) technique.;Traditionally, the discrete Fourier transform of a digital signal has been employed as a signal decomposition technique, as well as a modal identification technique by picking the peaks from its Fourier spectrum. However, the purposes and concepts of signal decomposition and modal identification are very different. While the performance of a signal decomposition technique would be judged based on the fitting between the reconstructed signal and the original signal, that of a modal identification technique could be judged based on whether identified modal parameters are close to the true modal parameters. When true modal parameters are unknown, the performance of a modal identification technique usually would be judged based on a stabilization diagram.;When a response signal, from either the tank test or the sea test, is modeled as the sum of many damped harmonic components, the numerical studies in this thesis demonstrates that using ERA to estimate component frequencies and damping ratios, together with a least-squares solution for getting amplitudes and phase angles, is an excellent signal decomposition technique. For modal identification, SSI is found to be better than ERA, and is a very efficient method for both the tank and the sea test data.;In their theoretical derivations, both ERA and SSI methods assume that the dynamic system is a time-invariant linear system. However, the real buoy-fluid system under investigation must be a nonlinear system, thus to apply ERA or SSI, a first approximation is to treat the dynamic system to be piecewise linear, i.e. linear within a short period. In this study, introducing a sliding window is for assuming that the system is linear within the window duration. With this sliding window, an ERA-based time-frequency analysis, in parallel to the short time Fourier transform (STFT), has been conducted. It was concluded that using ERA-based analysis could overcome the frequency resolution and leakage problems.