Machine learning-based Bayesian optimization of tuned inerter dampers for enhanced seismic response control in high-rise base-isolated structures

A Bayesian optimization-based design framework for Tuned Inerter Dampers (TIDs) is presented in this paper. The TIDs are used to high-rise base-isolated buildings that are subjected to ground motions that are far-fault, near-fault non-pulse, and near-fault pulse. The suggested approach clearly takes into account long-period dynamic characteristics and inertance-oriented parameterization, in contrast to traditional analytical tuning techniques. To effectively determine the ideal inertance, frequency ratio, and damping ratio that minimize seismic response metrics, a probabilistic Gaussian-process surrogate model is utilized. The optimized TID typically reduces mean square displacement by 20-25% under far-fault and near-fault non-pulse excitations and by 10-18% under near-fault pulse-type ground motions, according to thorough numerical studies performed on 30-, 40-, and 50-story base-isolated structures. The reduction of peak absolute accelerations at the top stories can reach 22.8%, which is a substantial improvement over traditional TMD, TMDI, and MTMDI systems. The suggested approach works best for mid-to high-rise buildings (30-40 stories), but because of higher-mode dominance, it eventually loses effectiveness for very tall structures. The current framework's primary drawbacks stem from its idealized inerter modeling and simplified single-degree-of-freedom representation, which may underestimate multi-modal and nonlinear effects in extremely tall buildings. However, under long-period and pulse-type earthquake excitations, the suggested Bayesian-optimized TID provides a reliable and computationally effective design method for enhancing the seismic resistance of high-rise base-isolated buildings.