A PROPOSAL FOR A BENCHMARK TO MONITOR THE PERFORMANCE OF DETAILED MONTE CARLO CALCULATION OF POWER DENSITIES IN A FULL SIZE REACTOR CORE

摘要

Current computers and Monte Carlo codes are not yet able to calculate the power density map in a full size reactor core with sufficient statistical accuracy. This is only expected in 10 to 20 years from now. To monitor the improvement in efficiency of Monte Carlo codes over the years a benchmark test is proposed with emphasis on the estimation of the power density in a large number of axial regions of each fuel pin. The modeling of the full size reactor core is kept simple but with sufficient geometrical detail for a realistic Monte Carlo calculation. The main aim of the benchmark is to determine the execution time for a Monte Carlo calculation of the power density with sufficient statistical accuracy in as many regions as possible and to monitor the efficiency of the calculation over the years considering improvements in Monte Carlo codes and computers. In this paper full specifications for geometry and composition are provided. Preliminary results with the MCNP5 code indicate that for a 1 % statistical uncertainty in the local power density more than 10~(10) neutron histories must be simulated. The computing time will increase substantially with the number of tallies for power density estimation in separate regions. For the time being this will limit the number of regions for which the power density can be estimated. There may also be memory problems when using very large numbers of tallies. It is expected that the number of available computer nodes and CPU cores for a parallel calculation will increase enormously in the coming decades. However, the noted decrease in efficiency when using more CPU cores will require more advanced programming of Monte Carlo simulation for parallel processing. Hence, the proposed benchmark will stimulate improvement in practical Monte Carlo calculations for reactor cores in many ways.