Trans. of KNS Autumn meeting , CD-ROM, 2009.
Monte Carlo methods are widely used because they are easy to implement and potentially very accurate. However, Monte Carlo methods have a drawback in high dominance ratio eigenvalue calculations (e.g., LWRs, NGNPs, etc.). When the dominance ratio becomes close to unity, the fission source distribution converges very slowly so that hundreds of cycles need to be skipped to obtain accurate results - regardless of how many particles are traced in each generation. A means of accelerating the source convergence of the distribution is strongly desired in such cases. Recently, the coarse mesh finite difference (CMFD) which was originally known as the nonlinear acceleration method was successfully applied to accelerate the source convergence in MOC calculations. This work is to examine the similar acceleration scheme to Monte Carlo calculations. Here, the CMFD linear system is obtained directly from MC tallies and the solution of the CMFD is used to adjust the source distribution of the MC calculation in the subsequent cycle. The performance of the proposed method is examined through a simple one-group high dominance ratio problem and also through a 2D multigroup problem. In an earlier work by Cho et al. so called p-CMFD rebalance was applied externally for the acceleration of MCNP calculations as an initial trial of CMFD acceleration. A different CMFD formulation and strategy are proposed here with a simplified internal multigroup MC code
