STRUCTURAL RELIABILITY ANALYSIS USING STOCHASTIC RESPONSE SURFACE METHOD

This paper aims to propose a stochastic response surface method considering correlated input random variables. The orthogonal transform is adopted to treat the correlated random variables in stochastic response surface method. Explicit polynomials are derived for the forth-order and fifth-order Hermite polynomial chaos expansions of random variables. A C#-language based computer program WHUSRSM (Wuhan University Stochastic Response Surface Method) is developed. Four examples are selected to illustrate the application of the proposed stochastic response surface method. The results indicate that the proposed stochastic response surface method can estimate the structural reliability involving correlated random variables efficiently. A third-order stochastic response surface method is reasonably accurate to calculate failure probabilities between 10-3 and 10-4. However, a fourth-order or fifth-order Hermite polynomial chaos expansions should be used for cases with high dependency between input random variables. The number of collocation points equaling twice the number of unknown coefficients does not ensure the accuracy. In general, it is recommended that the number of collocation points should be at least two times of the number of unknown coefficients of the Hermite polynomial chaos expansion.