Machine Learning-Based Correlation Analysis of Seismic Intensity Measures and Structural Response for Reinforced Concrete Frame Structures

Abstract

The correlation between seismic intensity measures (IMs) and structural response is crucial in accurately assessing the seismic performance of reinforced concrete (RC) frame structures. However, the correlation analysis using either nonlinear dynamic analysis (NDA) or common machine learning (ML) methods is time-consuming. To address these challenges, this paper proposes an ML-based method to accelerate the correlation analysis process using the recently developed low-rank matrix guided least-squares support vector machines for regression (LRLS-SVMR). Given a large-scale seismic response data set, the proposed method first employs LRLS-SVMR to extract features with a lower dimension. Then, a parametric prediction model can be learned from the data set, where the inputs are the extracted features and the output is the structural response quantity. The computational complexity of the learning process is only proportional to the dimension of the features and irrespective of the number of training data points. The proposed method is applied to the correlation analysis between IMs and structural response for three typical RC frame structures, considering different damage levels due to longitudinal reinforcement buckling. The prediction accuracy and computational efficiency of the proposed method are validated by comparisons with other widely used ML methods based on a large-scale data set covering 30,219 data points generated by the NDAs on the mentioned RC frame structures. The research findings show that the peak ground velocity correlates best with the maximum interstory drift ratio (MIDR), followed by spectra acceleration at the first period, peak ground displacement, and peak ground acceleration. Moreover, the structural damage due to longitudinal reinforcement buckling has a relatively small impact on the correlation between the mentioned IMs and the MIDR.

RIGHTS & PERMISSIONS

2025 Tianjin University and John Wiley & Sons Australia, Ltd.