Multi-parameter input damage limit prediction for bare reinforced concrete frames using XGBoost

Rapid and accurate seismic damage assessment of buildings is vital for pre-earthquake mitigation planning and efficient post-earthquake rescue operations. Inter-story drift ratio (IDR) damage limits are widely used in the seismic damage assessment of reinforced concrete (RC) frame structures. However, IDR limits for different damage levels are influenced by multiple key design parameters, such as axial load ratio, steel grade, and column slenderness. Relying on fixed IDR limits can cause prediction errors when evaluating RC frame structures with varying design parameters. This study introduces a machine learning-based method to predict IDR damage limits of bare RC frame structures by considering the impact of various design parameters. Subsequently, the influence of key design parameters on the damage limits of bare RC frame structures was analyzed. Finally, a dataset comprising 17,820 nonlinear time-history analyses of 162 bare RC frames with varying design parameters was established for model training and testing. Using the extreme gradient boosting (XGBoost) algorithm, the proposed method achieved 90.75 % prediction accuracy for damage states, outperforming the method based on fixed IDR limits. This study offers a valuable reference for the seismic damage evaluation of bare RC frame structures with varying design parameters. The findings have the potential to facilitate post-earthquake risk assessments and inform the development of emergency response strategies.