SSI identification and GA-based linear calibration of the isolation stiffness of an FPS-isolated school building: verification on synthetic ambient-like data

Seismic isolation is widely adopted to protect buildings and critical facilities by decoupling their response from ground motion and concentrating nonlinear energy dissipation within the isolation layer. Friction pendulum systems (FPS) provide self-centering and energy dissipation capabilities governed by a small set of uncertain mechanical parameters. A previous parametric study by the authors on the same FPS isolated school building showed that the high speed friction coefficient is the dominant nonlinear parameter, whereas the linear stiffness is the only parameter with a significant influence in the linear regime. A subsequent work introduced a two tier framework for nonlinear identification of friction parameters, based on a detailed three dimensional finite element model (FEM) and a condensed low order representation. However, this framework explicitly assumed that the linear stiffness properties of the isolation system had already been calibrated from ambient-vibration data. This paper addresses this missing step by presenting a preliminary procedure that combines stochastic subspace identification (SSI) of synthetic ambient-like vibrations with genetic-algorithm-based model updating of the isolation stiffness on the full finite element model. The genetic algorithm (GA) treats the model as a black box, updates the family-wise stiffnesses of the four FPS device types, and minimizes a cost function based on discrepancies in the first ten natural frequencies and the corresponding mode shapes, all condensed onto eight monitored horizontal degrees of freedom. Numerical results on synthetic data show that the proposed identification technique can define an updated model that reproduces the target modal properties with small frequency errors and high mode-shape correlation across these ten modes. This confirms that the linear stiffness of the isolators can be consistently recovered. The study paves the way for applying the same procedure to real structures, starting from the school building considered here, for which continuous monitoring data are already available.