Hybrid Weight Window Techniques for Time-Dependent Monte Carlo Neutronics

TL;DR

This paper introduces an automated hybrid weight window method for time-dependent Monte Carlo neutronics that improves variance reduction and spatial particle distribution accuracy by solving a low-order second-moment problem with noise filtering.

Contribution

The paper presents a novel hybrid weight window technique based on a low-order second-moment solution for efficient variance reduction in time-dependent Monte Carlo simulations.

Findings

Achieved uniform particle distribution in space.

Enhanced resolution of wave fronts and low-flux regions.

Validated on AZURV1 benchmark.

Abstract

Efficient variance reduction of Monte Carlo simulations is desirable to avoid wasting computational resources. This paper presents an automated weight window algorithm for solving time-dependent particle transport problems. The weight window centers are defined by a hybrid forward solution of the discretized low-order second moment (LOSM) problem. The second-moment (SM) functionals defining the closure for the LOSM equations are computed by Monte Carlo solution. A filtering algorithm is applied to reduce noise in the SM functionals. The LOSM equations are discretized with first- and second-order time integration methods. We present numerical results of the AZURV1 benchmark. The hybrid weight windows lead to a uniform distribution of Monte Carlo particles in space. This causes a more accurate resolution of wave fronts and regions with relatively low flux.