Search

This study investigates applying machine learning (ML) algorithms to estimate seismic fragility curves of reinforced concrete moment-resisting frames. The goal is to achieve computationally efficient estimations while maintaining accuracy compared to traditional methods. Various multi-classification algorithms are explored for this purpose. A greedy search method identifies key fragility features (structure period, soil shear wave velocity, ground motion intensity). Analysis shows that the decision tree method strikes an optimal balance between accuracy and speed. ML-based fragility curves closely align with analytical results, with area ratios under 10%, mean absolute error (MAE) below 0.06, and root mean square error (RMSE) below 0.082. [ABSTRACT FROM AUTHOR]