Conditional reliability analysis in high dimensions based on controlled mixture importance sampling and information reuse

In many contexts, it is of interest to assess the impact selected parameters on failure probability a physical system. To this end, one can perform conditional reliability analysis, in which becomes function these parameters. Computing requires recomputing probabilities for sample sequence parameters, strongly increases already high computational cost conventional analysis. We alleviate costs by reusing information from previous computations each subsequent analysis sequence. The method designed using two variants importance sampling and performs transfer densities analyses current one. put forward criterion selecting most informative densities, robust with respect input space dimension, use recently proposed density mixture model constructing effective dimensions. controls estimator coefficient variation achieve prescribed accuracy. demonstrate its performance means engineering examples featuring number pitfall features such as strong non-linearity, dimensionality small (10−5to10−9).