Rare event estimation using stochastic spectral embedding

Estimating the probability of rare failure events is an essential step in reliability assessment engineering systems. Computing this for complex non-linear systems challenging, and has recently spurred development active-learning methods. These methods approximate limit-state function (LSF) using surrogate models trained with a sequentially enriched set model evaluations. A proposed method called stochastic spectral embedding (SSE) aims to improve local approximation accuracy global, modelling techniques by residual expansions subdomains input space. In work we apply SSE LSF, giving rise embedding-based (SSER) method. The resulting partition space decomposes into easy-to-compute conditional probabilities. We propose modifications that tailor algorithm efficiently solve event estimation problems. include specialized refinement domain selection, partitioning enrichment strategies. showcase performance on four benchmark problems various dimensionality complexity LSF.