Fast Bayesian frequency domain modal identification from seismic response data

Abstract

Modal identification aims at identifying the modal parameters that mainly include natural frequencies, damping ratios and mode shapes. They provide baseline modal properties of objective structures and play an important role in the seismic design, structural health monitoring, model updating, etc. In existing monitoring systems, the input and output responses are usually recorded simultaneously, which allows the identification of the modal parameters using earthquake records in a short period of time. The Bayesian method can properly account for the uncertainty in accordance with probability logic for modal identification. This paper proposes a novel Bayesian method for modal identification using collected data during earthquakes with known input. The probability density function of modal parameters based on the Fast Fourier Transform of measured data is derived analytically. A fast algorithm has been developed to efficiently optimize the modal parameters, where the cases of closely-spaced modes and well-separated mode are applicable, even for a large number of measured degrees of freedom. A synthetic example and shaking table test were used to illustrate the efficiency and accuracy of the proposed method. Finally, this method was applied to a seven-story building - Van Nuys Hotel to investigate its dynamic characteristics using seismic response data.