Machine learning-driven performance-based seismic design of hybrid self-centering braced frames with SMA braces and viscous dampers

A hybrid self-centering braced frame equipped with shape memory alloy-based self-centering braces (SMA-SCBs) and viscous dampers is proposed to achieve enhanced seismic performance. Based on the proposed hybrid strategy combining the contributions of SMA-SCBs and viscous dampers, this paper investigates the advantages of such hybrid self-centering braced frames in achieving the desired maximum inter-story drift under a considered seismic intensity. To this end, the influence of design parameters of SMA-SCBs and viscous dampers on hybrid self-centering braced frames is examined through parametric dynamic analyses of equivalent single-degree-of-freedom systems. The analysis results indicate that the post-yield stiffness ratio αand energy dissipation factor βof SMA-SCB, and the contribution of viscous damper show significant influence on the peak displacement responses of hybrid self-centering braced frames. The constant inelastic displacement ratio prediction model for hybrid self-centering braced frames is developed using machine learning algorithms. A performance-based seismic design method is subsequently proposed for the hybrid self-centering braced frames to achieve the target displacement responses based on the developed machine learning models. Two hybrid self-centering braced frames are designed based on the proposed design method. Nonlinear dynamic analyses are conducted to investigate the seismic performance of the designed structures. To highlight the advantages of the hybrid self-centering braced frames, another six-story self-centering braced frame with SMA-SCBs only is also studied in this paper. The analysis results indicate that all the designed hybrid self-centering braced frames can achieve the desired performance objective. Compared to the self-centering braced frame with SMA-SCBs only, the hybrid self-centering braced frames can achieve much smaller base shear demand and absolute floor acceleration responses.