Sub-pin level distribution tallies and statistical error estimation with POD tallies in two-dimensional C5G7 benchmark

Kondo, Ryoichi   ; Yamamoto, Akio*; Endo, Tomohiro*

The flux distribution tallies using proper orthogonal decomposition, called POD tallies, have been developed to address issues of high-fidelity Monte Carlo simulations. The POD tallies can reduce both dimensionality and statistical error. The present study verifies the applicability of the POD tallies to sub-pin level flux distribution in the two-dimensional C5G7 benchmark. Statistical error estimation is also proposed by applying the circular block bootstrap method to the POD tallies to estimate the statistical error of the flux distribution in a single Monte Carlo calculation. In the verification, the dimensionality of the finely discretized distribution is reduced by more than 90% compared with conventional cell tallies. The statistical error is reduced by more than half as the average value of all tally regions. The proposed approach is confirmed to properly estimate the statistical error of flux distribution considering both the inter-cycle correlation and the correlation between the expansion coefficients of different POD orders.