Multi-objective optimization and deep learning-based design of base-isolated structures with tuned inerter dampers under near-fault ground motions

Near-fault pulse-like ground motions (NPGMs) can significantly amplify the seismic response of base-isolated structures (BIS). Achieving optimal simultaneous control of both displacement and acceleration in BIS under varying ground motions remains a persistent challenge. This paper investigates the seismic performance of BIS with tuned inerter damper (BIS-TID) subjected to NPGMs, and proposes an efficient, intelligent inverse design methodology grounded in deep learning. The novelty lies in using NSGA-II for multi-objective optimization on the simplified 3DOF BIS-TID model under analytical NPGM action, and then training a deep neural network to establish a direct mapping from structural parameters to optimal TID parameters, thereby enabling rapid and efficient inverse design. Unlike previous work, this approach explicitly considers pulse characteristics across diverse seismic scenarios, avoiding the high computational cost of conventional iterative methods. This enhances both scenario adaptability and computational efficiency, facilitating practical engineering implementation. A multi-degree-of-freedom (MDOF) building model is adopted as a numerical case study, wherein the isolation parameters are optimized and dynamic analyses are conducted using real NPGM records. The results demonstrate that the proposed optimization framework can effectively reduce the structural seismic responses and improve the control performance of the BIS. It bridges a critical research gap in the optimization design of isolation systems with additional inerter under NPGMs and intelligent optimization strategies in different seismic scenarios.