Qin, Jianjun3; Nishijima, Kazuyoshi1; Faber, Michael Havbro1
1 Department of Civil Engineering, Technical University of Denmark2 Section for Structural Engineering, Department of Civil Engineering, Technical University of Denmark3 Swiss Federal Institute of Technology
A New Scheme
The present paper presents a new scheme for probability integral solution for system reliability analysis, which takes basis in the approaches by Naess et al. (2009) and Bucher (2009). The idea is to evaluate the probability integral by extrapolation, based on a sequence of MC approximations of integrals with scaled domains. The performance of this class of approximation depends on the approach applied for the scaling and the functional form utilized for the extrapolation. A scheme for this task is derived here taking basis in the theory of asymptotic solutions to multinormal probability integrals. The scheme is extended so that it can be applied to cases where the asymptotic property may not be valid and/or the random variables are not normally distributed. The performance of the scheme is investigated by four principal series and parallel systems and some practical examples. The results indicate that the proposed scheme is efficient and adds to generality for this class of approximations for probability integrals.
Advances in Structural Engineering, 2012, Vol 15, Issue 11
System reliability; Scaled probability integrals; Asymptotic Laplace integral solutions; Monte Carlo simulations; Meta-level representation; Extrapolation scheme