Global reliability sensitivity analysis index and its efficient numerical simulation solution in presence of both random and interval hybrid uncertainty

Suting Zhou, Zhenzhou Lu, Chunyan Ling, Wanying Yun

Research output: Contribution to journalArticlepeer-review

8 Scopus citations

Abstract

In the presence of both random and interval hybrid uncertainty (RI-HU), investigating global reliability sensitivity (GRS) can identify the effect of random input on the structural safety globally. To this end, this work establishes the GRS index model and its corresponding efficient solution, and the innovation includes three aspects. Firstly, the GRS of the random input is defined by the average absolute difference between the failure probability upper bound (FP-UB) and the conditional FP-UB on the fixed random input under the RI-HU, and it can reflect the effect of the fixed random input on the structural safety. Secondly, the conditional FP-UB on fixing the random input at the realization is approximated by the conditional FP-UB on fixing the random input in a differential interval. Then the conditional probability theorem can be employed to convert estimating the GRS into the state recognition of the samples, which is the by-product of estimating the FP-UB by the numerical simulation. Finally, by the strategy of surrogating the performance function twice, the meta-importance sampling method is developed to efficiently estimate the GRS in presence of the RI-HU. The rationality of the proposed GRS index model and the efficiency of the developed estimation method are fully verified by the numerical and engineering examples.

Original languageEnglish
Pages (from-to)551-573
Number of pages23
JournalStructural and Multidisciplinary Optimization
Volume63
Issue number2
DOIs
StatePublished - Feb 2021

Keywords

  • Adaptive Kriging
  • Global reliability sensitivity
  • Hybrid uncertainty
  • Meta-importance sampling

Fingerprint

Dive into the research topics of 'Global reliability sensitivity analysis index and its efficient numerical simulation solution in presence of both random and interval hybrid uncertainty'. Together they form a unique fingerprint.

Cite this