Summary: | 碩士 === 國立成功大學 === 土木工程學系碩博士班 === 97 === Abstract
When a structural system is attacked by an earthquake, the variations of dynamic parameters, for example, the story lateral stiffness and the floor eccentric distance, could be identified in order to assess the damage severity and location. A nonlinear regression analysis in the time domain is performed in this study to identify the story parameters of torsion-coupled systems. Even though noises are added, the identified story lateral stiffness in strong-motion state is good enough to observe the damage severity for cases of full measurements. Initial guesses are required in performing the nonlinear regression analysis. These initial values of parameters are found here by using the equations of motion. If the response signals, which are numerically integrated through accelerograms with noises, are filtered with the help of wavelet transform and then are used in the equations of motion, then the calculated parameters would be better initial guesses in the following identification. Such a procedure gives very good results parametric of story lateral stiffness, which are validated by using both the responses of simulating systems and the records of several benchmark models under different base excitations.
|