Title: A neural-network-based variance decomposition sensitivity analysis

Authors: Francesco Cadini, Enrico Zio, Francesco Di Maio, Vytis Kopustinskas, Rolandas Urbonas

Addresses: Department of Nuclear Engineering, Polytechnic of Milan, Via Ponzio 34/3, 20133 Milan, Italy. ' Department of Nuclear Engineering, Polytechnic of Milan, Via Ponzio 34/3, 20133 Milan, Italy. ' Department of Nuclear Engineering, Polytechnic of Milan, Via Ponzio 34/3, 20133 Milan, Italy. ' Lithuanian Energy Institute, Breslaujos, str. 3, LT-44403 Kaunas, Lithuania ' Lithuanian Energy Institute, Breslaujos, str. 3, LT-44403 Kaunas, Lithuania

Abstract: This paper describes the implementation of an artificial neural network for obtaining the numerous model output calculations within a variance decomposition scheme for performing the model sensitivity analysis with respect to both individual and grouped parameters. A case study concerning the identification of the input variables mostly contributing to model output uncertainty, with reference to an accident scenario in a nuclear power plant, provides evidence of the effectiveness of the approach.

Keywords: neural networks; nuclear safety; RBMK-1500 accident scenario; sensitivity analysis; variance decomposition; nuclear energy; nuclear power plants.

DOI: 10.1504/IJNKM.2007.013565

International Journal of Nuclear Knowledge Management, 2007 Vol.2 No.3, pp.299 - 312

Published online: 06 May 2007 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article