An efficient algorithm for time-dependent failure credibility by combining adaptive single-loop Kriging model with fuzzy simulation

The time-dependent failure credibility (TDFC) can reasonably measure the safety level of the time-dependent structure under the fuzzy uncertainty, but the direct optimization algorithm to estimate the TDFC requires large computational cost and even results in locally optimal solutions. Therefore, an efficient method is proposed for estimating the TDFC by combining the fuzzy simulation and the single-loop Kriging model. In the proposed method, fuzzy inverse transformation theorem is firstly used to transform the estimation of the TDFC into a sample classification problem, in which the candidate sample pool generated by fuzzy simulation (FS) is classified into the failure group and the safety one. For improving the efficiency of the classification, a Kriging model is adaptively trained by an elaborate U-learning function in the candidate sample pool. After the candidate sample is divided into the failure group and the safety one by the convergent Kriging model, the TDFC can be estimated as a byproduct easily. The innovation of the proposed method includes two aspects: establishing the idea of the fuzzy simulation combined with the single-loop Kriging model to estimate TDFC efficiently and robustly, and designing an elaborate U-learning function to improve the efficiency of training the single-loop Kriging model. The presented examples validate the efficiency of the proposed method under the acceptable precision.