City Research Online

Abstract

A machine-learning surrogate model is developed to predict the hysteretic behavior of post-earthquake damaged RC columns based on a simulation framework. A comprehensive database including 1053 RC column specimens under cyclic loading is established. For each column, its hysteretic response is simulated using the modified Ibarra-Medina-Krawinkler (I-M-K) model. Based on the calibrated hysteretic models, a two-stage strategy is proposed to simulate the hysteretic responses of the damaged columns: (1) a partial loading history from the experimental data is first applied to the intact columns, leading to various levels of initial damage; and (2) the complete loading history is subsequently applied to the damaged columns. The backbone and cyclic deterioration parameters of the modified I-M-K hysteretic model are then identified from the simulated responses of the damaged columns. The backbone parameters identified for damaged columns are compared with those of the intact columns to quantify damage-induced reduction factors (DRFs). Six machine learning models are developed to predict the DRFs and cyclic deterioration parameters through the initial damage and column design parameters. The machine learning models are trained and tested using 1000 random train-test splits, and the performance is evaluated. It is found that the XGBoost model exhibits the best performance in predicting the hysteretic behavior of the damaged columns. Additionally, three RC column specimens outside the considered database are adopted to examine the general applicability of the developed XGBoost model. Finally, SHapley Additive exPlanations (SHAP) is used for interpretation, indicating that initial damage is the most influential feature.

Publication Type:	Article
Additional Information:	© 2026, The Authors. This is the accepted manuscript of an article published by Springer. Please refer to the publisher’s terms and conditions for information on re-use.
Publisher Keywords:	Hysteretic behavior, Machine learning, RC columns, Post-earthquake damage, Residual performance
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science T Technology > TA Engineering (General). Civil engineering (General) T Technology > TH Building construction
Departments:	School of Science & Technology School of Science & Technology > Department of Engineering
SWORD Depositor:	Symplectic Administrator

Export

Downloads