Study Of Monte Carlo Variance Reducing Techniques And Its Application In Fusion Reactor

There is a growing need to perform neutronics calculations for China Fusion Engineering Test Reactor(CFETR).It includes computationally intensive,high resolution radiation transport and shutdown dose analyses.In the large and complex reactor geometries,since shielding material increased attenuation of neutrons,very few of the starting source particles leave the vacuum vessel to contribute to mesh elements where far from the plasma source.Variance reduction method is a solution that bias the analog transport probabilities could allow more samples be taken in the regions of interest.For the neutron deep penetration problem,using the one-dimensional fusion reactor(CFETR)slab model,geometry splitting and Russian roulette,forced collision,weight window were experimented and analyzed.Typical problems of variance reduction application were reasoned and summarized,like particle flat distribution problem,all zeros in weight window problem.Some new thoughts of variance reduction technique for deep penetration shielding problem were introduced.By different combinations of variance reduction techniques,the combination of weight window and exponential transform performed well that achieved 162 times than the analogue simulation.However,the local variance reduction can hardly handle the variance reduction problem for a global system.Global variance reduction(GVR)method is an efficient solution for global Monte Carlo particle transport problems.The method uses neutron information to construct the biasing parameters.In this paper,a GVR procedure was developed and applied.Cylinder CFETR model was used for comparing various forms of GVR method.Results showed that flux based GVR method ensured more reliable statistical results in everywhere of the model,which achieved an efficiency of 7.43 times than AN.Mesh tally of the neutron flux was chosen for the GVR method to simulate global neutron transport of large three-dimensional CFETR model.By ten GVR weight window iterations,non-analog particles of the system were distributed uniformly and sampled adequately.All voxels were scored and the average relative error was 5.7%in the ultimate step of GVR iteration.When the GVR weight window is implemented in large and complex models,streaming path and deep penetration could lead to a long history problem that takes a disproportionate amount of time for the accomplishment of Monte Carlo simulations,and exerts a detrimental effect on the efficiency of parallel computing.In this research,the limitation of weight window splitting was proposed.If the total number of weight window splitting reaches a certain value,the weight window splitting behavior would be terminated.MCNP was modified to revise the weight window function.Tests were conducted on a streaming dog-leg geometry model and a 3D model of the CFETR.The results show that a suitable FWWS parameter in the alleviation method could optimize both the performance of variance reduction and the efficiency of parallel calculation.which makes long history problem tractable.