The Study of a New Efficient Monte Carlo Method for Deep-Penetration Transport

Abstract

Comparing with deterministic methods, Monte Carlo method has high precision but huge time-consuming when using for shielding design. For real deep-penetration problems, a series of variance reduction methods have been proposed and applied in related software (e.g. MCMP, SERPENT) in recent decades to overcome the drawbacks of Monte Carlo method. However, these methods still have troubles, such as the selection of correction factors and function model in biasing method. The important region division method also has time and memory consuming issues in complicated models. At present, the Consistent Adjoint-Driven Importance Sampling (CADIS) and Forward-Weighted CADIS (FW-CADIS) methods are implemented well in deeply penetrating problems. This paper presents a new efficient Monte Carlo method to solve deep-penetration problems. Contrary to traditional Monte Carlo methods, in this method, the particle trajectories that contributes to the tallies most are first determined, then the occurrence probability of the corresponding trajectory is calculated and counted. The pre-determined tracks are obtained through a serious of geometric transformations from standard tracks generated in a simple medium. The geometric transformations of tracks include rotation and stretching/shortening. Moreover, the weight correction is performed to assure the weight is unbiased. Preliminary numerical results on monolayer medium demonstrate that this method can significantly reduce calculation consumptions while retaining decent accuracies.