The Monte Carlo method is widely used in neutron transport calculations, especially in complex geometry and continuous-energy problems. However, an extended application of the Monte Carlo method to large realistic eigenvalue problems remains a challenge due to its slow convergence and large fluctuations in the fission source distribution.
In this thesis, a deterministic partial current-based Coarse-Mesh Finite Difference (p-CMFD) method is proposed that achieves fast convergence in fission source distribution in Monte Carlo k-eigenvalue problems. In this method, the high-order Monte Carlo method provides homogenized and condensed cross section parameters while the low-order deterministic p-CMFD method provides anchoring of the fission source distribution.
The proposed method is implemented in the MCNP5 code (version 1.30) by appending a scattering cross section tally routine based on a collision estimator and a deterministic p-CMFD acceleration routine. The new method is tested not only on multigroup problems, but also on realistic one- and two-dimensional heterogeneous continuous-energy large core problems. For the problems tested, the Monte Carlo anchoring method using p-CMFD shows much faster convergence in the eigenvalue and fission source distributions (in both of the required number of inactive generations and total computation time) compared to the conventional MCNP5 code due to the accelerated anchoring distributions.
For realistic continuous-energy problems, the Monte Carlo anchoring method using “dynamic” anchoring factor ($alpha$ =1.0 during inactive generations and $alpha$ =0.0 during active generations) and a cross section accumulation scheme shows noticeably improved results in spite of stochastic fluctuation in generated cross sections. The numerical results demonstrated that the Monte Carlo anchoring method accelerates realistic continuous- energy, heterogeneous k-eigenvalue Monte Carlo problems quite well.