Reliability Analysis Method Based on Pre-Interpolation-Active Learning Kriging

In view of the shortcomings of the traditional structural reliability analysis method based on Kriging surrogate model in local optimization and interpolation calculation, a pre-interpolation reliability analysis method based on improved active learning strategy (PI-AK-MCS+U) was proposed. In this method, the intermediate process point and intermediate process Kriging model are innovatively introduced to construct a dual optimization search mechanism, which realizes optimization iteration in the process of each round of reliability analysis model update, and significantly reduces the computing cost. The active learning function is improved by multiple punishment mechanisms, and the sample points near the limit state surface are adaptively selected to avoid local aggregation of sample points. A convergence criterion based on the stability of the failure probability is developed to improve the local prediction accuracy of the model. The two-dimensional lowprobability failure problem, the multi-failure domain tandem system and the six-dimensional nonlinear oscillator system are used as examples to verify the problem. The results show that compared with the traditional method, the time required for the proposed method and the number of iterations required to accurately assess the probability of failure are significantly reduced.