An effective single loop Kriging surrogate method combing sequential stratified sampling for structural time-dependent reliability analysis

Introducing surrogate models in time-dependent reliability analysis is an effective means to reduce the computational burden, but improving the efficiency of reliability analysis is still a challenge. In this paper, an effective single loop Kriging surrogate method combing stratified sampling for structural time-dependent reliability analysis is developed to improve the computational efficiency of time-dependent reliability. First, by dividing the candidate samples into four regions, reliability analysis is performed by sequential sampling in each region. In addition, the candidate samples for each analysis are smaller than the initial set of generated original samples, which can improve the computational efficiency of time-dependent reliability. Second, in order to select the candidate samples that contribute much to the reliability analysis, a new learning function is proposed that can consider the misclassification probability and probability density of the candidate samples. Finally, in order to further accelerate the convergence of the algorithm, a combined convergence criterion is proposed that is able to consider the relative error of failure probability, coefficient of variation, and convergence state. The accuracy and effectiveness of the method are demonstrated by four examples.