Machine learning‐aided resilience‐based seismic design of RC frame structures

As the focus of structural engineering shifts from collapse prevention to post‐earthquake recovery, the assessment of seismic resilience has become increasingly crucial. Currently, the field of seismic design lacks methods that directly incorporate resilience, primarily due to their computational complexity and time‐consuming nature. To address this gap, this study proposes an innovative seismic design approach for reinforced concrete (RC) frame structures aimed at optimizing resilience indicators. In the proposed design framework, seismic resilience indicators are explicitly considered during the preliminary design stage of RC frame structures. First, a comprehensive database for evaluating the seismic resilience of RC frames is established using the OpenSees simulation platform and the PACT resilience assessment tool. Subsequently, a machine learning–based predictive model is developed to estimate the resilience indicators of RC frame structures. The relationships between design parameters and resilience indicators are interpreted through SHapley Additive exPlanations and sensitivity analysis. Finally, an optimization framework integrating the Random Forest algorithm, resilience indicators, and a multi‐objective particle swarm optimization algorithm is established. Finally, three RC frame structures with different numbers of stories were designed to validate the effectiveness of the proposed method. The results indicate that beam span, story height, beam section depth and longitudinal reinforcement ratio, as well as column section depth and longitudinal reinforcement ratio, have significant effects on seismic resilience indicators. After optimization, the life‐cycle cost of the three‐story frame structure decreased by 20.73%, injuries were reduced by 7.06%, and fatalities decreased by 14.29%.