Development of a Hybrid Deterministic-Stochastic Method for Full Core Neutronics

Research efforts in reactor physics focus on the improvement of current analysis methods for neutronics or development of advanced ones. Recently, there is a renewed interest in stochastic simulations due to their superior accuracy and the advances in computing platforms. In particular, there is an interest in the development of hybrid stochastic deterministic methods for accelerating the inactive cycles of Monte Carlo. However, such hybrid approaches are not very efficient as the inactive cycle constitute a very small portion of the simulation. In this work, a novel hybrid method is developed and discussed. The approach combines conventional Monte Carlo with deterministic flux mapping to reduce computational costs of full core simulations. The main contributors to the computational expenses of Monte Carlo is the number of tallies scored and the number of neutron histories tracked. In the proposed hybrid method, the Monte Carlo simulation is performed with a small number of neutron histories while maintaining good confidence and scoring flux tallies on a coarse mesh. Then,the dominant modes of the transport equation and flux mapping are employed to reconstruct the neutron flux and reaction rates on a finer mesh. Application to a number of example problems show that the studied hybrid method can achieve up to 90% reduction in the computational time compared to conventional Monte Carlo while maintaining comparable accuracy.

Résumé

de l'étude de l'équation de transport et de l'approximation des probabilités de collision est présenté. En plus de la solution du mode fondamental décrivant la répartition asymptotique du flux neutronique, l'équation de transport suppose un grand nombre de solutions qui peuvent être très utiles dans les études de la perturbation, la cartographie et la synthèse des flux neutroniques.