Comparison of Sampling Schemes in Asymptotic Sampling

Magdalena Smidova¹; Miroslav Vorechovsky¹

¹ Brno University of Technology

doi:10.20906/CPS/USM-2016-0071

Resumo

This article deals with the possibility to use Asymptotic Sampling (AS) for estimation of the failure probability. The AS algorithm requires samples of multidimensional Gaussian random vector. There are many alternatives how to obtain such a sample and selection of the sampling strategy influences the performance of AS method. Three reliability problems (testing functions) are selected to test AS. First, the functions are analyzed using AS in combination with Monte Carlo designs and LHS designs optimized using Periodic Audze-Eglājs (PAE) Criterion. Afterwards, the same set of problems has been solved without AS procedure by direct estimation of failure probability. All the results are also compared with the exact value of the failure probability.

Palavras-chave: Asymptotic Sampling; Failure Probability; Monte Carlo (MC); Latin Hypercube Sampling (LHS); Periodic Audze-Eglajs (PAE)