Structural time and space dependent reliability analysis based on active learning surrogate model
The limit state function of structures is often implicit and influenced by uncertainty,time,and space,rendering its output a time-dependent random field.Structural failure is typically a small probability event,requiring numerous finite element simulations,making existing methods difficult to apply.First-order/second-order reliability analysis methods are available,but their accuracy and efficiency are limited.This paper proposes a time and space-dependent reliability analysis method based on the active learning surrogate model,aimed at enhancing efficiency.Firstly,derive the probability of extreme value sign prediction for a random sample in the time and space domains and then construct a corresponding learning function.Selecting candidate samples with the minimum sign prediction probability to add to the training set and sequentially updating the Kriging model until the accurate prediction probability of all candidate samples exceeds 99%,the failure probability is computed simultaneously.The efficiency and accuracy of the proposed method are validated through three case studies,offering a new way to time and space reliability analysis.
time and space dependent reliabilitysurrogate modelactive learningKriging modelfailure probability
万方数据
维普
