Study on Recursive Subspace Identification of State-Space Models

Abstract

The purpose of this thesis is to apply the multivariable recursive subspace identification (RSI) algorithm to the computer-aided analysis program By obtaining the acceleration responses of the finite element models during earthquake excitation as input and output data for system identification to continuously identify the system parameters through recursive algorithm The identified system parameters and the known system mass matrix are used to estimate the stiffness matrix of the system and compare the identification and estimation results with that computed by the finite element model to verify the accuracy and applicability of this system identification method In this thesis several finite element models with different shapes and degrees of freedom are established and artificial earthquakes with different peak ground accelerations (PGA) and different dominant periods (Ts) are applied The RSI method is used to continuously identify the system parameters of these simulation cases during seismic excitation In addition in reality the effect of noise on acceleration measurement must be considered so white noise with different noise-to-signal ratios is applied to simulate the influence of the noise on system identification and the tolerance of noise of RSI method can be known The results of this thesis indicate that this RSI method can obtain good identified results in various structural models and under different types of earthquake excitations In addition it also shows that the RSI method can maintain respectable identification results when the amplitude of the noise is not too large Note that the computer-aided analysis programs in this thesis are developed by Shen-Haw Ju’s research team which are free to use

Date of Award	2020
Original language	English
Supervisor	Shen-Haw Ju (Supervisor)