Using Neural Networks as Nonlinear Model Interpolators in Variance Decomposition-based Sensitivity Analysis
In this paper, we investigate the possibility of embedding neural networks, appropriately trained on the results of a Monte Carlo plant reliability evaluation, within a classical decomposition scheme for efficiently performing multiparametric sensitivity analyses of a reliability model. These analyses are of great importance for the identification of critical systems, structures and components of hazardous plants, such as nuclear or chemical ones, thus providing significant insights for their risk-based design and management.
Bibliographic Reference: An oral report given at: SAMO 2001: Third International Symposium on Sensativity Analysis of Model Output. Organised by: CIEMAT. Held in: Madrid Spain, 18-20 June, 2001
Record Number: 200013410 / Last updated on: 2001-06-27
Original language: en
Available languages: en