中文摘要：

如何更高效发挥减方差的作用是蒙特卡罗方法在先进核能系统屏蔽分析应用中的研究热点之一。本文发展了一种基于自动网格划分与权窗平滑的自适应减方差方法，在蒙特卡罗和确定论耦合的一致性伴随驱动的重要性抽样方法（Consistent Adjoint Driven Importance Sampling，CADIS）基础上，利用计算机辅助设计（Computer Aided Design，CAD）自动转换和自由程自动划分曲网格，通过确定论方法伴随预计算，实现基于伴随通量的区域权窗参数自动配置，并对伴随通量变化剧烈区域进行权窗平滑处理，保证了粒子在不同区域的有效偏倚，进一步提高计算的效率，从而解决大空间蒙特卡罗计算难以收敛的问题。该方法已初步应用于中国铅基反应堆（China Lead．based Reactor,CLEAR）堆顶盖的屏蔽计算分析，该案例具有结构复杂、屏蔽材料厚重的特点，测试结果表明本方法将计算效率提高近10倍。

英文摘要：

Background： In the radiation shielding analysis of advanced nuclear reactor with Monte Carlo method, it is one of the researches focuses on using variance reduction methods more effectively. Purpose： An adaptive variance reduction method based on automatic mesh generating and weight window smoothing is developed in this paper. Methods： With hybrid calculation of Monte Carlo method and deterministic theory, consistent adjoint driven importance sampling （CADIS） method was developed by automatic mesh generating with computer aided design （CAD） and pre-calculation of free path. The adjoint flux calculation results with deterministic method are treated as core values to generate source biasing parameters and weight window parameters in the form of mesh. And smooth treatment is introduced for alleviating the lower bound of weight window violent change to make sure the efficiency of particle bias in different areas and improve calculation efficiency. This method can solve the convergence problem in Monte Carlo simulation with large space geometry. Results： The adaptive method has been applied in the shielding calculation of China lead-based reactor （CLEAR） cover with complex geometry and massive shielding body. The result showed that the computing efficiency is improved by nearly 10 times. Conclusion： The adaptive variance reduction method can be applied to the shielding calculation of advanced nuclear fission reactor.