Statistical system identification of structures using ARMA models
Conte, Joel P.
Master of Science
The present study focuses on the estimation of the modal properties, i.e., natural modal frequencies, modal damping ratios, and mode shapes, of a structural system under ambient vibration based on the response data of the structure and the statistical properties of the stationary loading process. The corresponding input-output system is represented by a covariance-equivalent ARMA model which is univariate in the case of multiple input - single output and multivariate in the case of multiple input - multiple output. The mapping between the physical modal parameters and the ARMA parameters is established. Two different algorithms are used to estimate the covariance-equivalent ARMA models, namely the maximum-likelihood algorithm and a two-stage least-squares algorithm. An augmented system approach including both a fictitious "load" system and a structural system is formulated to account for the non-whiteness of most realistic loading processes. Examples of its application are provided for the cases of earthquake, wave, and wind excitations.
Civil engineering; Electronics; Electrical engineering