Efficient Algorithm for Estimating Generalized Failure Probability Function of Aeroengine Turboshafts

For the strength reliability model of an aeroengine turboshaft that sufficiently considers the random inputs and fuzzy state, the generalized failure probability function (G-FPF) quantifies structural reliability impact from varying distribution parameters (DP) of random inputs. However, directly estimating G-FPF with a double-loop framework is time-consuming. To alleviate this issue, an efficient single-loop stochastic collocation method is proposed by sufficiently employing the characteristics of the turboshaft performance function. Firstly, by separating strength and stress variables in the performance function of an aeroengine turboshaft, G-FPF is equivalently transformed into the one-dimensional reduced integral of a continuous integrand function, on which the G-FPF can be directly estimated by the efficient stochastic collocation method. Secondly, a unified density weight function, independent of DP but enveloping their design domains, is constructed to estimate the transformed G-FPF in the space of a one-dimensional reduction. Then, a set of stochastic collocation points of the unified density weight function can be used to estimate the G-FPF, eliminating the dependence of the computational cost on the number of DP realizations. Results show that the proposed method is efficient under acceptable accuracy, and its computational cost is independent of the DP dimensionality and the size of its design regions.