2018

Häkkinen, Silja

Variance reduction routines are often employed when making particle transport calculations with Monte Carlo codes in a geometry involving regions of very low particle density. This improves the statistical accuracy of the calculation and reduces the calculation time. Shielding calculations are a type of particle transport calculation where the particle density is intentionally kept low in the regions of interest and the use of variance reduction is profitable. <br/><br/>Recently, variance reduction techniques in the form of weight windows have been implemented in VTT’s Monte Carlo code Serpent. In this work, Serpent’s variance reduction methods have been tested by calculating photon dose rate and neutron flux on the surface of a castor storage cask in air. The cask was filled with burned UO2 or MOX assemblies and the figure of merit, FOM, of calculations with and without variance reduction was compared.<br/><br/>Variance reduction improved FOM in the photon transport calculations by several orders of magnitude compared to calculations without variance reduction. In the neutron transport calculations, FOM was approximately 3 times higher with variance reduction than without it. This is because the castor storage cask absorbs neutrons far less effectively than photons. The calculated photon dose rates and neutron flux with variance reduction corresponded to the values calculated without variance reduction within relative standard deviation in most cases. In the cases where the two values differed from each other, it was concluded to be caused mainly by the poor statistics of the calculations without variance reduction. Some of these cases without variance reduction were calculated with a larger number of particle histories and the differences vanished.