Kernel Density Estimation Method in Time-Dependent Monte Carlo Simulation

Abstract

With the development of computer technology, direct Monte Carlo simulation of reactor transient behaviors has received more attention. The state-of-the-art method in time-dependent Monte Carlo simulation involves linearizing the equations over a time step; snapshots of the neutrons at the time boundaries are created as the source for the next time step. How to construct the neutron distribution at time boundary is an important new problem. In this paper we present a kernel density estimation (KDE) method in sampling time boundary sources. This method can obtain the continuous distribution from discrete time boundary sources. For subcritical systems, KDE method effectively improves confidence of direct Monte Carlo simulation.