An asymptotic stochastic response surface approach to reliability assessment under multi-source heterogeneous uncertainties

• Published in: Reliability engineering & system safety Vol. 215; p. 107804
• Main Authors: He, Jingjing, Huang, Min, Wang, Wei, Wang, Shaohua, Guan, Xuefei
• Format: Journal Article
• Language: English
• Published: Barking Elsevier Ltd 01.11.2021; Elsevier BV
• Subjects: Adequacy; Approximation; Asymptotic methods; Asymptotic properties; Asymptotic reliability assessment; Coupling; Curvic coupling; Design of experiments; Fatigue life; Low cycle fatigue; Materials fatigue; Mathematical models; Model accuracy; Monte Carlo simulation; Multi-source heterogeneous uncertainties; Numerical models; Reliability analysis; Reliability engineering; Response surface methodology; Stochastic response surface; Structural reliability; Uncertainty; Variance decomposition; Asymptotic reliability assessment; Variance decomposition; Stochastic response surface; Curvic coupling; Multi-source heterogeneous uncertainties
• ISSN: 0951-8320; 1879-0836
• DOI: 10.1016/j.ress.2021.107804

An asymptotic stochastic response surface method for structural reliability assessment under multi-source heterogeneous uncertainties is proposed in this study. The fatigue reliability of an aeroengine Curvic coupling under the uncertainties from material, contact surface state, and part dimension deviation is used to motivate the methodology development. A randomized block design of experiments for numerical models are employed to obtain the fatigue strain responses under different variations and combinations of the uncertain variables. A stochastic response surface model is constructed, and the analytical forms of the mean and bound predictions of the fatigue strain are derived. By further coupling the fatigue strain into the low-cycle fatigue model, the fatigue life under multi-source uncertainties is obtained. The stochastic moment approximation is used to derive the asymptotic approximation of the fatigue life distribution, allowing for efficient reliability and risk-informed lifetime prediction. The variance of the fatigue life contributed by individual uncertain sources is naturally resolved in the method, providing a quantitative measure for uncertainty control and management. The adequacy of the stochastic response surface model and the accuracy of the asymptotic reliability assessment result are verified using statistical testing and the Monte Carlo method, respectively.