Improvement of Frequency Domain Output Only Modal Identification from the Application of the Random Decrement Technique

This paper explores the idea of estimating the spectral densities as the Fourier transform of the random decrement functions for the application of frequency domain output-only modal identification methods. The gains in relation to the usual procedure of computing the spectral densities directly from the time series, are due to the noise reduction that results from the time averaging procedure of the random decrement technique, and from avoiding leakage in the spectral densities, as long as the random decrement functions are evaluated with sufficient time length to have a complete decay within that length. The idea is applied in the analysis of ambient vibration data collected in a ¼ scale model of a 4-story building. The results show that a considerable improvement is achieved, in terms of noise reduction in the spectral density functions and corresponding quality of the frequency domain modal identification results.This paper explores the idea of estimating the spectral densities as the Fourier transform of the random decrement functions for the application of frequency domain output-only modal identification methods. The gains in relation to the usual procedure of computing the spectral densities directly from the time series, are due to the noise reduction that results from the time averaging procedure of the random decrement technique, and from avoiding leakage in the spectral densities, as long as the random decrement functions are evaluated with sufficient time length to have a complete decay within that length. The idea is applied in the analysis of ambient vibration data collected in a ¼ scale model of a 4-story building. The results show that a considerable improvement is achieved, in terms of noise reduction in the spectral density functions and corresponding quality of the frequency domain modal identification results.