Prediction of and genetic algorithm optimization on data induced uncertainty reduction with the use of an integral experiment

Abstract

This paper verifies a method for predicting the uncertainty reduction on a parameter of interest in a target application system from adjusting nuclear data based on an integral experiment and developed a procedure for optimizing integral experiments for particular applications. This paper also discusses the expansion of a global optimization software package, Gnowee, to optimize the design of the Flexible Neutron Source (FNS) at the University of Tennessee Knoxville, in combination with MCNP, for the purpose of reducing uncertainties from nuclear data on applications’ parameters of interest. The optimization process considered various isotopes and reactions, and the final design reduced the number of sodium plates almost to the minimum allowed while using a combination of fuel plates and empty spaces. The final design is predicted to be able to reduce the cross section induced uncertainty on a generic sodium fast reactor by 15–20 % on the k-eigenvalue.