THE TWO-PHASE TRAINING STRATEGY AND HIGHLY-EFFICIENT RELIABILITY ANALYSIS METHOD

To enhance the accuracy and efficiency of constructing adaptive meta-models, a two-phase adaptive training strategy is proposed. On the basis of the two-phase training strategy, a class of efficient reliability analysis methods is further developed by combining the probability density evolution method. In order to verify the effectiveness of the proposed method in this paper, the construction of meta-models with different types of functional functions as objectives is analyzed, and the seismic reliability analysis of a reinforced concrete frame structure is conducted. The results show that the two-phase adaptive training strategy greatly improves the export rate of the target agent model and retains high analysis accuracy, which makes up for the lack of accuracy of the probability density evolution method in dealing with rare failure events.