Acceleration of source convergence in Monte Carlo criticality calculation with CMFD schemes

The CMFD acceleration schemes have been demonstrated to effectively accelerate the fission source convergence of Monte Carlo (MC) k-eigenvalue problems. In this paper, the performance of the lpCMFD scheme implemented with MC was investigated and compared to the standard CMFD and pCMFD schemes. Two test problems were developed for this study, a simple one-dimensional monoenergetic problem and a two-dimensional assembly problem. The Shannon entropy was analyzed for the test cases to evaluate the convergence rate of fission source. To reduce stochastic errors of the CMFD parameters between cycles, an accumulation scheme was implemented to average over the previous cycles. It is found that averaging of the 5 previous cycles yields the best convergence performance, and it also helps to improve the stability of CMFD. The real variance and inter-iteration correlation were analyzed for the lpCMFD scheme at different averaging lengths. Numerical results show that lpCMFD performs best. It is also demonstrated that the successive over-relaxation (SOR) factor can further improve the acceleration performance of lpCMFD for MC source iteration. In addition, it is shown that lpCMFD can mitigate the clustering phenomena that occurs in MC calculations. (authors)