A structural seismic reliability prediction method based on adaptive sampling and Gaussian process regression

This study aims to develop an efficient and accurate method for predicting the seismic collapse reliability of structures. To address the high computational cost associated with traditional sampling-based methods for structural reliability quantification, an adaptive Gaussian Process Regression (GPR) framework is proposed. This framework employs an adaptive sampling strategy with a two-step global and local optimization process to enhance predictive accuracy. The proposed approach is validated on a typical four-story reinforced concrete (RC) frame and benchmarked against traditional sampling methods. The results demonstrate that the adaptive GPR framework effectively captures the nonlinear structural response and exhibits excellent convergence and accuracy in predicting collapse probability. Compared to traditional Monte Carlo simulation (MCS), the proposed method significantly reduces computational costs, further highlighting the importance of considering parameter uncertainty in seismic performance assessment. This study provides a reliable engineering tool for the design and seismic performance evaluation of large-scale and complex structures.