Back to home
A new efficient sampling method for quantifying and propagating nuclear data uncertainty in CUSA   HAO Chen1, LI Peijun1, ZHAO Qiang1, and ZHOU Haining2   1. Fundamental Science on Nuclear Safety and Simulation Technology Laboratory, Harbin Engineering University, Harbin, China (haochen.heu@163.com) 2. The University of Michigan, Ann Arbor, MI, US   Abstract: Uncertainty inevitably exists in nuclear data and correlation exists between different nuclear reactions, which are always represented by the relative covariance matrix. For the statistical sampling method, the major technical challenge is to generate a desirable input sample space by using an efficient sampling method based on the relative covariance matrix. In this paper, an efficient sampling method of Latin Hypercube Sampling combined with Singular Value Decomposition Conversion (LHS-SVDC) is proposed based on rigorous mathematical derivation and especially the correlation information between different cross sections is represented precisely. TMI-1 pin-cell case of OECD UAM benchmark was employed to verify this new method with respected to the reference solution generated by TSUNAMI-1D module. The numerical results indicate that the new LHS-SVDC method can generate a desirable sample space of multi-group cross sections quickly and effectively, which can further propagate the uncertainty in multi-group cross sections to the target parameters more accurately. Keywords: nuclear data; uncertainty; correlation; efficient sampling method; SVD
	Click here for viewing full text.