A novel surrogate-model based active learning method for structural reliability analysis

The surrogate-model based active learning method has a satisfactory trade-off between efficiency and accuracy, which has been widely used in reliability analysis. In this paper, an active learning function called the potential risk function (PRF) is proposed to adaptively estimate the failure probability. It should be emphasized that the proposed potential risk function is not limited to the Kriging metamodel, which can be combined with other mainstream surrogate models in principle. Further, an effective convergence based on the failure probabilities in 10 consecutive iterations is adopted to prevent the pre-mature of the surrogate-model based active learning method (SM-ALM). Four validation examples (one numerical example, two benchmark examples, and one practical engineering problem) are applied to validate the robustness and effectiveness of the proposed SM-ALM. •Based on the surrogate model, a novel active learning method is proposed for structural reliability analysis.•A novel active learning function can strengthen the fitting precision near the limit state surface.•An effective convergence criterion is adopted to terminate the active learning process.•The proposed method can be applied to any existing surrogate models in principle.•The proposed method can balance the computational efficiency and accuracy well.