Improved Non-Intrusive Polynomial Chaos for Reliability Analysis under Hybrid Uncertainty

With the increasing of systems' scale and complexity, reliability analysis faces more challenges which mainly include hybrid uncertainty, implicit limit state function and numerous uncertain input variables. Non-intrusive polynomial chaos (NIPC) is a promising technology for uncertainty quantification with high efficiency and accuracy. However, as polynomial chaos is defined in probability space, NIPC is not applicable to reliability analysis under hybrid uncertainty with multiple input variables. To address this issue, an improved NIPC approach is proposed that Klir log-scale transformation is employed to unify fuzzy variables and random variables. And a combinatorial optimization algorithm is developed to efficiently select the optimal collocation points for NIPC with multiple uncertain inputs. Comparative study on the airborne retractable system shows that the proposed approach can achieve higher accuracy than response surface method with identical computational cost.