Application of Adaptive Multilevel Splitting on Coupled Neutron-Photon TRIPOLI-4 Monte Carlo Simulations - Archive ouverte HAL Access content directly
Conference Papers Year : 2018

Application of Adaptive Multilevel Splitting on Coupled Neutron-Photon TRIPOLI-4 Monte Carlo Simulations

(1) , (2)
1
2

Abstract

In the context of radiation protection simulations, the Adaptive Multilevel Splitting (AMS) algorithm is a challenging variance reduction (VR) technique that has been recently investigated in the field of particle transport simulation. It has been implemented in the forthcoming version 11 of the Monte Carlo code TRIPOLI-4® and successfully tested in neutron-only and photon-only configurations. This paper addresses the application of the AMS algorithm to coupled simulations, and particularly to neutron-photon Monte Carlo calculations. The branching process occurring during the Monte Carlo coupled transport is taken into account in the new coupled-AMS algorithm and is explained in this paper. Two different neutron-photon configurations are then investigated, leading to a comparison of the coupled-AMS algorithm with the analog simulation on the one hand, and with the Exponential Transform (ET) on the other hand, which is the standard VR technique of TRIPOLI-4. Gains up to 30 are obtained in terms of Figure of Merit relatively to the analog simulation, which is about 4 to 6 times more efficient than the ET method for these configurations.
Fichier principal
Vignette du fichier
201800001997.pdf (201.29 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

cea-02338605 , version 1 (21-02-2020)

Identifiers

  • HAL Id : cea-02338605 , version 1

Cite

Henri Louvin, Odile Petit. Application of Adaptive Multilevel Splitting on Coupled Neutron-Photon TRIPOLI-4 Monte Carlo Simulations. RPSD-2018 - 20th Topical Meeting of the Radiation Protection & Shielding Division of ANS, Aug 2018, Santa Fe, United States. ⟨cea-02338605⟩
19 View
23 Download

Share

Gmail Facebook Twitter LinkedIn More