Prediction model and sensitivity analysis of ultimate drift ratio for rectangular reinforced concrete columns failed in flexural-shear based on BP-Garson algorithm

Abstract

Considering that the deformation capacity of reinforced concrete (RC) columns is crucial in performance-based seismic design, the ultimate drift ratio of RC columns failed in flexural-shear are studied using BP neural network and Garson (BP-Garson) algorithm. Firstly, this study employed data onto tests conducted on 35 rectangular RC columns with flexural-shear failure. Nine influencing factors including axial load ratio, shear span ratio, longitudinal reinforcement configuration, transverse reinforcement configuration, and material strength were selected as variables, a BP neural network was used to establish the relationship between the ultimate drift ratio and these variables. Accordingly, a predictive model for the ultimate drift ratio of RC rectangular columns failed in flexural-shear was constructed. Then, the Garson algorithm for local sensitivity analysis was applied to determine the sensitivity of each influence of the ultimate drift ratio of RC rectangular columns failed in flexural-shear. The results showed that the BP neural network could accurately predict the ultimate drift ratio of RC rectangular columns with flexural-shear failure, of which the primary influencing factors are shear span ratio, axial load ratio, volume transverse reinforcement ratio and longitudinal reinforcement ratio. The BP-Garson algorithm integrates model development with influence analysis, which could improve computational efficiency and serves as a reference for evaluating the sensitivity of seismic performance and its influencing factors in other structural components.