Pin to pin neutron flux reconstruction in a PWR reactor using support vector regression (SVR) technique

Abstract

Coarse mesh nodal methods are widely used in the analysis of nuclear reactors. However, these methods provide only average values of the neutron fluxes. From a safety point of view, it is important to have an accurate analysis of the pin to pin flux distribution that nodal methods are not able to provide. Many articles have been published that make use of mathematical techniques to determine flux distributions. Most of these techniques use expansion functions to estimate these distributions. The expansion coefficients of these works are determined by conditions that take into account the average values of certain fluxes supplied by the nodal methods. There are also methods that employ analytical solutions of the neutron diffusion equation. This article presents a different approach for calculating the pin to pin neutron flux distribution for a PWR reactor. The developed method uses support vector regression (SVR) technique to determine this pin to pin neutron flux. The SVR technique uses average data computed with the Nodal Expansion Method (NEM) for learning purposes. A total of 70% of the computed data were used for training and 30% for validation, using multifold-cross-validation. Two fuel elements were removed from the training and validation sets, to test the method. Less than 2% errors were found when compared to the values ​​obtained by the nodal expansion method (NEM), using a fine-mesh spatial discretization. We concluded that use of SVR to reconstruct pin to pin fluxes is another option, which will be of great value in fuel reload calculations, since the same parameters will be applied to all cycles, thus expediting calculations when compared to standard procedure calculations.