American Nuclear Society

Home / Publications / Journals / Nuclear Science and Engineering / Volume 149 / Number 1

Stationarity Modeling and Informatics-Based Diagnostics in Monte Carlo Criticality Calculations

Taro Ueki, Forrest B. Brown

Nuclear Science and Engineering / Volume 149 / Number 1 / January 2005 / Pages 38-50

Technical Paper /

In Monte Carlo criticality calculations, source error propagation through the stationary (active) cycles and source convergence in the settling (inactive) cycles are both dominated by the dominance ratio (DR) of fission kernels. For symmetric two-fissile-component systems with the DR close to unity, the extinction of fission source sites can occur in one of the components even when the initial source is symmetric and the number of histories per cycle is more than 1000. When such a system is made slightly asymmetric, the neutron effective multiplication factor at the inactive cycles does not reflect the convergence to stationary source distribution. To overcome this problem, relative entropy has been applied to a slightly asymmetric two-fissile-component problem with a DR of 0.993. The numerical results are mostly satisfactory but also show the possibility of the occasional occurrence of unnecessarily strict stationarity diagnostics. Therefore, a criterion is defined based on the concept of data compression limit in information theory. Numerical results for a pressurized water reactor fuel storage facility with a DR of 0.994 strongly support the efficacy of relative entropy in both the posterior and progressive stationarity diagnostics.

Questions or comments about the site? Contact the ANS Webmaster.