In this work, a dynamic identification methodology in the absence of real experimental data is presented, aimed at developing diagnostic tools for the post-earthquake assessment of seismic isolation system performance. The methodology has been specifically developed for an existing base-isolated building with pendulum-type seismic isolators where a monitoring system has recently been installed. This approach will allow identification of the global nonlinear hysteretic behavior of the isolators by analyzing structural response data acquired during strong-motion seismic events.

The structure is represented using a simplified three-degree-of-freedom (3DOF) model, where the displacements of the points immediately below and above the isolators and those at the roof level are considered as key dynamic variables. The total isolator stiffness is obtained by summing the stiffness values of the individual isolators in parallel, while the interstory stiffness is determined via static analyses of a comprehensive finite element model. The masses are defined by aggregating the participating masses of the relevant floors. The column stiffness, calculated as the sum of the stiffnesses of the columns, completes the mechanical characterization of the system. In the absence of monitoring data, this 3DOF model is utilized to generate synthetic structural responses, which are employed as a substitute for experimental data. An iterative optimization algorithm calibrates the global isolator parameters (stiffness, friction coefficient) by minimizing discrepancies between theoretical performance targets and simulated responses. The real seismic accelerograms used as inputs for numerical simulations serve a dual purpose: training the parameter identification process through synthetic data assimilation and testing the model’s predictive capability under unseen ground motions. The approach will thus enable the detection of significant parameter variations, which may indicate mechanical degradation, structural modifications, or modeling inaccuracies.

This work establishes the basis for future developments, including multi-degree-of-freedom models with torsional effects and the integration of machine learning techniques for advanced rheological characterization and prediction of response under extreme loading, ultimately enabling comprehensive diagnostic systems for the post-seismic management of base-isolated structures.