Title: Reliability assessment using stochastic local regression
Authors: Seung-Kyum Choi
Addresses: The George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Savannah, GA 31407, USA
Abstract: A primary challenge of stochastic analysis is to discover rigorous ways to estimate the low probability of failure which is critical to reliability constraints. In this paper, a new framework is proposed for the improved estimation of the low failure probability. Combining the significant advantages of the polynomial chaos expansion, Karhunen-Loeve transform and local regression method will result in a new simulation-based modelling technique that enables the accuracy of the structural integrity prediction. The proposed procedure can allow for realistic modelling of sophisticated statistical variations and facilitate in order to achieve improved reliability by eliminating unnecessary conservative approximations. Several specific examples including a three-bar truss and an unmanned undersea vehicle are depicted to illustrate how the method is used to provide a quantitative basis for developing robust designs associated with the low probability of failure.
Keywords: polynomial chaos expansion; Karhunen-Loeve transform; local regression; moving least-squares; reliability assessment; uncertainty; stochastic analysis; low failure probability; simulation; modelling; structural integrity prediction; three-bar truss; unmanned vehicles; undersea vehicles.
International Journal of Reliability and Safety, 2009 Vol.3 No.1/2/3, pp.267 - 285
Published online: 27 Jun 2009 *Full-text access for editors Access for subscribers Purchase this article Comment on this article