Answers are grounded in this paper and cite their sources.
Maintaining high prediction accuracy in structural reliability analysis under limited computational budgets and a limited number of high‐cost function evaluations remains a significant challenge. To address this issue, an active learning Gaussian process regression (AGPR) method is proposed for reliability assessment. The proposed method improves the quality and spatial distribution of the initial training samples, enabling more effective use of limited computational resources and reducing the influence of redundant samples on surrogate model construction. Furthermore, the surrogate model is adaptively refined in critical regions associated with failure probability estimation, allowing more accurate characterization of the limit‐state boundary with a limited number of function evaluations. By improving the representativeness of training samples and refining the surrogate model near the limit‐state boundary, the proposed method can provide accurate failure probability estimates with a limited number of function evaluations. Numerical and engineering examples indicate that the proposed method can achieve accurate failure probability estimates with fewer function evaluations than several existing approaches in the considered examples. These results indicate the potential of the proposed method for structural reliability analysis under limited computational budgets.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: