Modal parameter estimation from ambient data using time-frequency analysis

Abstract Modal parameter estimation of randomly and ambient excited structures using output-only system identification techniques has become a recently essential issue for assessment of engineering structures, structural health monitoring and structural control. A number of mathematical models on the output-only identification techniques have been developed in the time domain, in the frequency domain and in the most recent time-frequency domains. Time-frequency analysis using wavelet transform advantages to represent measured response in the simultaneous time and frequency domains which facilitate for estimating both natural frequencies in the frequency domain and damping in the time domain. However, the output-only identification using the time-frequency analysis have difficulties such as time-frequency resolution analysis for damping estimation and close frequencies, especially for extracting high-order and low-energy modal parameters and so on. This paper presents theoretical bases of the wavelet transform for output-only system identification of the ambient data with emphasis on extracting the natural frequencies and damping. Full-scale measurements have been carried on a five-storey steel structure. Some refinement techniques using discrete wavelet decomposition with multi-resolution analysis, empirical mode decomposition as well as using broadband and narrowband filtering are also proposed for estimating the modal parameters of high-order mode and close-frequency problems. Modified complex Morlet wavelet, furthermore, is used for flexible adaptation on the time and frequency resolutions.