基于共轭离散纵标的减方差方法

对于深穿透类型的屏蔽计算，为了得到较为可信的统计结果，蒙特卡罗方法(MC方法)需要模拟大量的粒子，巨大的计算时间是其存在的主要问题。源偏倚和权窗技巧能够有效降低深穿透问题的计数误差。开展了基于共轭离散纵标(SN)的MC减方差方法研究，根据SN方法的共轭注量率计算并生成了源偏倚和权窗参数，编写了JMCT程序的源抽样子程序，并且在秦山一期测量值基础上进行了验证，成功应用到CAP1400压力容器快中子注量率和堆腔中子和光子剂量率计算中。数值结果表明，对于深穿透屏蔽计算问题，和无偏的MC方法相比，基于共轭SN的MC减方差方法能够在保证结果精度的前提下，提高计算效率1~2个量级。

Variance reduction method based on adjoint discrete ordinate

For deep-penetration shielding calculation, Monte Carlo method (MC method) requires modeling a great number of particles to obtain reliable results, thus huge computation time is the main problem of the MC method. Source biasing and weight window technique effectively decrease the tally error of deep penetration problem. This paper studies the variance reduction (VR) method based on adjoint Discrete Ordinate (SN), generates source biasing factors and weight window parameters for the MC method by using the adjoint fluence rates of the SN method, and develops source sampling subroutine for JMCT. The VR method was verified at phaseⅠ of Qinshan Nuclear Power Plant measurements. It was applied to CAP1400 pressure vessel fast neutron fluence rate and cavity neutron and photon dose rate calculations. Numerical results show that the VR method based on SN increases calculation efficiency by 1~2 orders for deep-penetration shielding calculation with high precision compared with unbiased MC method.