Identification of first and second order models for the superstructure of base-isolated buildings using free vibration tests: A case study

Using a set of full-scale free vibration measurements, two separate state of the art identification strategies are implemented for the time domain identification of the modal properties of the superstructure of a three-story base-isolated building. In a first approach, the superstructure is modeled as a fixed-base linear 3-DOF system subjected to the base acceleration measured during the tests. Identification is performed using the Covariance Matrix Adaptation Evolution Strategy (CMA-ES), a stochastic algorithm for non-linear, black-box optimization. In the second approach, a first order state-space representation of the superstructure is considered, and identification is carried out using the Eigensystem Realization Algorithm with Data Correlation, complemented by the Observer/Kalman filter Identification algorithm (ERA-DC/OKID). A single-input and multiple-output system is considered, the input being the recorded base acceleration and the output given by the measured floor accelerations. The results of the two procedures are compared leading to considerations about modeling assumptions that are typically adopted in the analysis and design of base-isolated buildings.