Automated acceleration and convergence testing for Monte Carlo NCS calculations

Monte Carlo methods have been used for over 60 years in nuclear criticality safety (NCS) calculations. Significant burdens are placed on NCS analysts to properly run the calculations: -1) The initial guess for fission sites is defined by user input; -2) Users must ensure that sufficient neutrons/cycle are used to prevent bias; and -3) Users must ensure that enough cycles are discarded so that k_eff and the fission source have converged. In practice, a short run produces plots of k_eff and entropy, then the number of inactive cycles is manually set in the MCNP6 input file, and a final run is made. NCS work often requires parameter studies with 100 s of runs. For these studies, it is not practical to follow all of the procedures above, and conservative over-estimates are used for the inactive cycles. Recent work has addressed these burdens, providing automated acceleration of the convergence process, statistical tests for automatically determining convergence, and additional tests to assess whether a sufficient number of neutrons/cycle was used. These automated methods do not require user input and provide quantitative evidence of convergence. Testing on a wide range of problems has demonstrated that the methods are robust and reliable. (authors)