Surrogate-assisted boundary-oriented multi-objective optimization of semi-rigid steel frame-brace-damper systems

Abstract

Brace-damper co-design can improve the seismic performance of steel frames, but current optimization practice is limited by three issues: semi-rigid connection flexibility is often neglected; conventional Pareto searches may represent the attainable performance envelope unevenly; and multi-record nonlinear time-history analysis (NLTHA) is too costly for direct use throughout evolutionary optimization. This study develops a boundary-oriented surrogate-assisted workflow for multi-objective design of semi-rigid steel frame-brace-damper systems. It combines dynamic EAA for nonlinear response evaluation, BNSGA-II for boundary-oriented search, and an MLP surrogate for candidate screening. In analytical benchmark tests, BNSGA-II increases the mean number of occupied boundary regions by 34.1% and 34.8% relative to NSGA-II in the bi- and tri-objective problems, respectively. In the six-story case, the cost-drift envelope identifies four distinct regions and a transition from brace-only control to mixed damping below f θ ≈ 0.004 rad. In the twenty-story case, tri-objective optimization reveals clear investment-performance trade-offs and identifies mixed-damper solutions that substantially suppress primary-member inelastic development while satisfying the adopted acceleration criterion. Computationally efficient surrogate screening, followed by NLTHA verification, provides a basis for brace-damper selection under different cost and performance targets.