This paper presents a statistically refined Bouc-Wen model of tri-axial interactions for the identification of structural systems under tri-directional seismic excitations. Through limited vibration measurements in th...This paper presents a statistically refined Bouc-Wen model of tri-axial interactions for the identification of structural systems under tri-directional seismic excitations. Through limited vibration measurements in the National Center for Research on Earthquake Engineering in Taiwan conducting model-based experiments, the 3-D Bouc-Wen model has been statistically and repetitively refined using the 95% confidence interval of the estimated structural parameters to determine their statistical significance in a multiple regression setting. When the parameters＇ confidence interval covers the ＂null＂ value, it is statistically sustainable to truncate such parameters. The remaining parameters will repetitively undergo such parameter sifting process for model refinement until all the parameters＇ statistical significance cannot be further improved. The effectiveness of the refined model has been shown considering the effects of sampling errors, of coupled restoring forces in tri-directions, and of the under-over-parameterization of structural systems. Sifted and estimated parameters such as the stiffness, and its corresponding natural frequency, resulting from the identification methodology developed in this study are carefully observed for system vibration control.展开更多
文摘This paper presents a statistically refined Bouc-Wen model of tri-axial interactions for the identification of structural systems under tri-directional seismic excitations. Through limited vibration measurements in the National Center for Research on Earthquake Engineering in Taiwan conducting model-based experiments, the 3-D Bouc-Wen model has been statistically and repetitively refined using the 95% confidence interval of the estimated structural parameters to determine their statistical significance in a multiple regression setting. When the parameters＇ confidence interval covers the ＂null＂ value, it is statistically sustainable to truncate such parameters. The remaining parameters will repetitively undergo such parameter sifting process for model refinement until all the parameters＇ statistical significance cannot be further improved. The effectiveness of the refined model has been shown considering the effects of sampling errors, of coupled restoring forces in tri-directions, and of the under-over-parameterization of structural systems. Sifted and estimated parameters such as the stiffness, and its corresponding natural frequency, resulting from the identification methodology developed in this study are carefully observed for system vibration control.
