Methods of Monte Carlo biasing using two-dimensional discrete ordinates adjoint flux
Methods of biasing three-dimensional deep penetration Monte Carlo calculations using importance functions obtained from a two-dimensional discrete ordinates adjoint calculation have been developed. The important distinction was made between the applications of the point value and the event value to alter the random walk in Monte Carlo analysis of radiation transport. The biasing techniques developed in this study are the angular probability biasing which alters the collision kernel using the point value as the importance function and the path length biasing which alters the transport kernel using the event value as the importance function. Source location biasing using the step importance function and the scalar adjoint flux obtained from the two-dimensional discrete ordinates adjoint calculation were also investigated.
The effects of the biasing techniques to Monte Carlo calculations have been investigated for neutron transport through a thick concrete shield with a penetrating duct. Source location biasing, angular probability biasing, and path length biasing were employed individually and in various combinations. Results of the biased Monte Carlo calculations were compared with the standard Monte Carlo and discrete ordinates calculations. Based upon the fractional standard deviations of the answers, the biasing techniques are a factor of 2 to 8 better than the standard method. Hence, the effectiveness and the applicability of the biasing techniques in deep penetration, three-dimensional Monte Carlo calculations are clearly demonstrated.
Results of Monte Carlo calculations using respectively the event value and the point value to alter the transport kernel showed that the event value calculation gave much better statistics (the FSD reduced by a factor of 2) than the point value calculation. Therefore, it was confirmed that the event value is the more appropriate function for biasing the transport kernel.
Thesis76b.T253.pdf
4.46 MB
Adobe PDF
ee905d83d2d2d7e0e7591e5de9ee96a1