The artificial neural network used in the study of sensitivities in the IRIS reactor pressurizer

The technique of sensibility analysis studies the behavior of the ratio between the variation of output results and the variation of input parameters in general. This study performed in the reactor pressurizer, which is a component responsible for controlling of the pressure inside the vessel, has the fundamental importance in designing the security of any concept of an advanced reactor. In fact, for its feature of passive action of the pressurizer (there is no spray), this analysis becomes a necessary step for safety and performance of the plant. The direct method through code MODPRESS, which represents the pressurizer model of the International Reactor Innovative and Secure (IRIS), has required a huge computational effort. To solve this problem, artificial neural networks (ANNs), beyond faster, has been used to replace the MODPRESS in this article. The ANNs do not require linear behavior of the system and can use both, simulated or experimental data for their training and learning. In order this, we adopted a classical non-supervised training ANN for mapping and forecasting of the pressurized using initially simulated data. In next future, we will incorporate the experimental data from the operation of the CRCN-NE reduced-scale test facility mapping. Moreover, based on the results obtained in this study, one can conclude that the artificial neural networks are presented as an alternative to MODPRESS code, and artificial neural networks are actually a great tool to calculate the sensitivity coefficient. (C) 2013 Elsevier Ltd. All rights reserved.