Document Type

Article

Abstract

In this paper, several alternative approaches are used to implement the failure sampling method for structural reliability analysis and are evaluated for effectiveness. Although no theoretical limitation exists as to the types of problems that failure sampling can solve, the method is most competitive for problems that cannot be accurately solved with reliability index-based approaches and for which simulation is needed. These problems tend to have non-smooth limit state boundaries or are otherwise highly nonlinear. Results from numerical integration and three extrapolation approaches using the generalized lambda distribution, Johnson's distribution, and generalized extreme value distribution are compared. A variety of benchmark limit state functions were considered for evaluation where the number of random variables, degree of non-linearity, and level of variance were varied. In addition, special limit state functions as well as two complex engineering problems requiring nonlinear finite element analysis for limit state function evaluation were considered. It was found that best results can be obtained when failure sampling is implemented with an extrapolation technique using Johnson's distribution, rather than with numerical integration or the generalized lambda distribution as originally proposed with the method.

Disciplines

Applied Mechanics | Structural Engineering

Comments

This is the final draft of an article published in ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering, 2(4), (2016) © ASCE, available online at: http://dx.doi.org/10.1061/AJRUA6.0000876

Download

Share

COinS