1. Seismic demand amplification of steel frames with SMAs induced by earthquake sequences.
- Author
-
Ke, Ke, Zhou, Xuhong, Zhu, Min, Yam, Michael C.H., and Zhang, Huanyang
- Subjects
- *
STEEL framing , *EARTHQUAKE aftershocks , *INDUCED seismicity , *SHAPE memory alloys , *MACHINE learning , *EARTHQUAKES , *STEEL analysis - Abstract
This paper investigates the inelastic seismic demands of steel frames equipped with shape memory alloys (SMAs) subjected to mainshock-aftershock earthquake sequences. Based on the multiple nonlinear stages of steel frames equipped with SMAs, a trilinear self-centring hysteretic model is introduced and validated first. Then, inelastic spectral analyses of steel frames equipped with SMAs subjected to mainshock-aftershock earthquake sequences are conducted. In particular, the energy modification factors of the corresponding single-degree-of-freedom (SDOF) under single mainshocks and mainshock-aftershock earthquake sequences are examined and compared. The results show that the energy modification factor of steel frames equipped with SMAs under mainshock-aftershock earthquake sequences is higher than that under single mainshocks, and the energy modification factor shows sensitivity to trilinear self-centring hysteretic parameters. Last, an amplification coefficient is developed and estimated for quantifying the amplification effect of recorded mainshock-aftershock earthquake sequences on the energy modification factor. Specifically, several machine learning (ML) algorithms are implemented and compared for estimation and interpretation. The results indicate that the XGBoost model performs best in predicting the amplification coefficient with the highest coefficient of determination of 0.9845. Besides, the interpretable ML approaches including partial dependence plot (PDP) and shapely additive explanations (SHAP) are proved helpful in explaining the trained ML model. • A trilinear hysteretic model is proposed and verified. • Effects of the influential parameters on energy modification factors are investigated. • Amplification effect of mainshock-aftershock earthquake sequences is quantified. • Prediction model for amplification effect is developed utilising machine learning technologies. • Interpretable machine learning approaches are adopted. [ABSTRACT FROM AUTHOR]
- Published
- 2023
- Full Text
- View/download PDF