Variance estimation and central limit theorem in Monte Carlo criticality calculation

Ueki, Taro  

A new methodology has been developed to make the reliable estimation of statistical errors in Monte Carlo criticality calculation (MCCC). The methodology developed is directly based on the convergence process in the functional central limit theorem and is shown to perform well in the evaluation of reactor power distribution. The theoretical backbones are described within the general context as framed in the operations research. The requisite basics of statistics are reviewed in terms of output analysis in MCCC. Numerical results are presented for the initial core model of a 1200 MWe pressurized water reactor. Preliminary results of fractal dimension analysis are shown to discuss a potential for convergence assessment.