全局减方差方法在大空间γ辐射场计算中的应用

Application of global variance reduction methods for the calculation ofγradiation field in a large space

为了提高大空间γ辐射场全局计算效率,开展了全局减方差(Global Variance Reduction,GVR)方法在大空间γ辐射场计算中的应用研究。针对计数栅元/网格体积差异造成的过度分裂问题,引入体积修正因子修改全空间权窗参数。体积修正后的基于通量的GVR方法计算的全局品质因子(FOMG)比直接模拟提高约39倍。针对非计数区计算耗时问题,提出了非计数区修正方法,使得FOMG因子进一步提高40%。在引入体积和非计数区修正的基础上,在大空间γ辐射场计算中与基于粒子误差、权重、径迹、数目、能量、碰撞和通量的7种GVR方法进行对比。结果表明:7种GVR方法计算的FOMG因子比直接模拟提高2~3个量级,基于误差的标准差σ降低2~3个量级;而基于权重的GVR方法计算的FOMG因子比直接模拟提高2304倍,在所有GVR方法中减方差效果最好。在基于通量的GVR方法中引入光滑因子SI后,模拟计算的权窗下限随SI增加而减小,FOMG因子随SI的增加先增加后减小。当SI=0.8时,该方法计算的FOMG因子最大,比直接模拟提高3246倍。

[Background]The direct simulation of theγradiation field in a large space has a very low calculation efficiency.[Purpose]This study aims to apply the global variance reduction(GVR)method to the calculation of theγradiation field in a large space.[Methods]Firstly,the volume correction factor was introduced for modifying the lower limit of the weight window wth to address the over-splitting problem caused by the volume difference between the counting cells/grids.The global quality factor(FOMG factor)calculated by the flux-based GVR method using the volume correction was 39 times higher than that obtained by direct simulation.Then,a non-counting area correction method was proposed to address the time-consuming problem encountered in non-counting area calculation while the FOMG factor was further improved by 40%.Finally,based on both the volume correction and non-counting area correction,the calculation of theγradiation field were compared with that of seven GVR methods based on the particle error,weight,track,number,energy,collision and flux,respectively.The smoothing factor SI was introduced into the flux-based GVR method for results further analysis.[Results and Conclusions]The results show that the FOMG factor calculated by the seven GVR methods is about 2~3 orders of magnitude higher than that obtained by direct simulation,and the standard deviationσis reduced by 2~3 orders of magnitude.The FOMG factor calculated by the weight-based GVR method is 2304 times higher than that obtained by direct simulation;this value yields the best variance reduction effect among all GVR methods.As SI increases,the lower limit of the weight window wth of the simulation decreases,and the FOMG factor first increases and then decreases.When SI=0.8,the calculated FOMG factor has the largest value,which is 3246 times higher than that obtained by direct simulation.