Conference paper Open Access
Bodi, Janos; Ponomarev, Alexander; Mikityuk, Konstantin
The paper presents a hybrid stochastic-deterministic method in which one-group cross sections are 1) generated with a Monte Carlo code for selected core states using the whole-core model and 2) adjusted for diffusion solution applying a sampling technique to improve the agreement between the prediction of static parameters by the Monte Carlo and nodal diffusion codes. The hybrid method was applied for an analysis of the Superphenix Sodium Fast Reactor core using Serpent 2 as a continuous energy Monte Carlo code and the PARCS (Purdue Advanced Reactor Core Simulator) reactor kinetics code as a nodal diffusion solver. Assuming correspondence between two models, the universes of the Serpent 2 whole-core model are defined to represent the nodes of the PARCS whole-core model. Universe-wise one-group cross sections and fluxes and k-effective value were generated for selected core states using Serpent 2. The cross sections were then used in a series of PARCS simulations involving a correction sampling technique, an iterative procedure, in which one-group transport cross sections are randomly variated assuming uniform distribution in a specified range in order to find a PARCS solution for k-effective and 3D flux distribution as close to the Serpent 2 results as possible. To obtain a more global picture about the capability of the hybrid method and its convergence, three core configurations have been simulated, namely, the reference critical core, a Doppler reactivity effect case with increased temperatures of fissile and fertile materials, and a sodium density reactivity effect with reduced sodium density. For all core configurations, a good agreement has been achieved between the PARCS and Serpent 2 predictions of the static parameters (k-effective within 10 pcm and flux within 2%), providing a potential for fast-running and acceptably accurate transient core simulations.