Structural reliability assessment through surrogate based importance sampling with dimension reduction

We present a method for reliability assessment in extreme conditions from numerical simulator through surrogate based importance sampling. As proposed recent works the literature, Kriging is used to build an approximation of limit state function and optimal density. Our contribution then use sufficient dimension reduction which enables construction metamodel lower dimension. The so called augmented failure probability correction factor are recast this framework. Simple strategies refinement subspace described and, case Gaussian inputs, computationally efficient MCMC scheme aimed at sampling quasi-optimal density presented. non-Gaussian inputs also laid out it argued demonstrated simulations that approach can reduce number calls computer model, crucial analysis. Advantages supported by carried on industrial study concerned with response prediction wind turbine under loading.