公開文献から読み取れるモンテカルロ計算における分散低減法の懸念事項

モンテカルロ計算の論理=経験から科学へ

抄録

In the fixed source problem such as a neutron deep penetration calculation with the Monte Carlo method, the application of the variance reduction method is most important for a high figure of merit (FOM) and the most reliable calculation. But, MCNP calculation inputs written in published literature are not to be best solution. The most concerned items are setting method for the lower weight bound on the weight window method and the exclusion radius for a point estimator. In those literatures, the lower weight bound is estimated by engineering judge or weight window generator in the MCNP. In the latter case, the lower weight bound is used with no turning process. Because of abnormal large lower weight bounds, many neutron are killed in no meaning by the Russian Roulette. The adjoint flux method for setting of lower weight bound should be adapted as a standard variance reduction method. The Monte Carlo calculation should be turned from the experience such as engineering judge to science such as adjoint method.