Variance reduction for fixed-source Monte Carlo calculations in multiplying systems by improving chain-length statistics

摘要

For a fixed-source calculation in a multiplying system, the variance in the neutron-chain length can create an unstable calculation and a neutron flux with high variance. In this paper an analytical frame is derived to calculate a priori the variance of the neutron flux in a simple model of a fixed-source calculation. This method is then used to find better variance reduction techniques not only for this simplified problem, but also for general cases. Two new variance reduction techniques are found: fully implicit fission and the branchless-collision method. These two new methods are compared to the existing variance reduction techniques present in MCNP5 and Tripoli4. It is shown that these new methods are more effective for variance reduction in a multiplying system than the methods presently used. For a test problem the Figure of Merit increased by more than a factor 10. The new variance reduction techniques are applicable in a general case.