Application of a global variance reduction method to HBR-2 benchmark

It is a great challenge to obtain reliable results in reasonable time by the Monte Carlo (MC) method in solving the deep-penetration shielding problem. Based on Discrete Ordinate (SN) fluence rate, local and global variance reduction (LVR and GVR) methods use the biased source and weight window technique to decrease MC calculation tally error for the deep-penetration problem. This paper analyses calculation efficiency of the LVR and GVR methods for the HBR-2 benchmark. Numerical results show that both the LVR and GVR methods obtain reliable results for the HBR-2 benchmark. The LVR method requires separate SN and MC calculation for each dosimeter; whereas the GVR method simultaneously optimizes both in-vessel and ex-vessel dosimeters using an adjoint source weighted by SN forward response. The application of the GVR method is more efficient and convenient than LVR method.