Uncertainty in the seismic performance of semi-active base isolation systems

In a conventional base isolation system, minimizing the seismic responses of the superstructure is always at the cost of increasing the isolator's response. The semi-active control of the isolator has been considered an effective solution to such a dilemma. It tunes the real-time properties of the isolator according to preset rules to further reduce the superstructure's seismic responses without increasing that of the isolator or vice versa. However, the number of ground motion records used to design and validate the controller, i.e., the preset rules, in existing studies is usually very small and therefore is suspectable if it is adequate to address the significant uncertainty in the shaking of future earthquakes. This paper critically reviews the performance of the proportional-integral-derivative (PID), linear-quadratic regulator (LQR), and fuzzy controllers in semi-active base isolation systems with magnetorheological (MR) dampers subjected to highly uncertain ground motion inputs through numerical simulations. The results show that the control performance of the controllers varies significantly with the increasing number of input records, suggesting the necessity of using at least 50 ground motion records to appropriately assess the performance uncertainty of semi-active base isolation systems. More importantly, the superior performance of the optimized controllers is not guaranteed if the system is subjected to ground motions that are new to the controller, even if the controller has been optimized for thousands of existing ground motions. It highlights the need of improving the adaptability of the semi-active systems for uncertain ground motion inputs.