Research On Nuclear Data Uncertainty Propagation Method

With the rapid development of computer simulation technology and programming models,nuclear data uncertainty is the major factor affecting reactor safety parameters. And in order to comply with the "high fidelity" requirement of nuclear reactor physics calculation,to improve the reliability of reactor physics results, it is of great significance to carry out the research on nuclear data uncertainty by using high-efficiency and reasonable uncertainty analysis methods. In the process of nuclear data uncertainty propagation, statistical methods and deterministic methods are effective. This paper focuses on the uncertainty analysis method based on the sampling statistics, and as a supplement and verification method of sampling statistical results, the sensitivity and uncertainty analysis is carried out by using direct numerical perturbation method.On the basis of traditional reactor physics calculation process, this paper established the framework of nuclear data uncertainty propagation, and identified the uncertain input and output of each physical calculation process. In order to reasonably quantify nuclear data uncertainty,this paper introduced the group transform method to produce multi-group nuclear cross-section covariance matrix and the self-consistency rules of nuclear data perturbation.Based on nuclear cross-section covariance matrix and the cross-section self-consistency rules,this paper established nuclear cross-section perturbation propagation model. At the same time,the paper studied the uncertainty and sensitivity analysis methods based on statistical sampling and direct numerical perturbation methods, which provided the theoretical methods for nuclear data uncertainty analysis.Due to the correlation existed between nuclear reaction cross sections, in order to effectively use statistical sampling methods to carry out nuclear data uncertainty analysis, this paper developed an efficient sampling method which combined latin hypercube sampling(LHS) and Cholesky decomposition transformation. After numerical verification, the efficient sampling method of LHS combined with Cholesky decomposition transform was reasonable and feasible, and it could quickly and efficiently obtain the random sample space. Based on the efficient sampling theory, nuclear reactor physics calculation uncertainty analysis program CUSA was upgraded to be suitable for nuclear data uncertainty analysis. The sensitivity results of the direct numerical perturbation method were compared with those of SCALE program, and the uncertainty results of the sampling statistics method were compared with those of the direct numerical perturbation method. The comparison results show that nuclear cross-section perturbation propagation model is correct, and the method and procedure of nuclear data uncertainty analysis based on efficient sampling theory are reasonable and feasible.Finally, nuclear data uncertainty analysis methods were applied to TMI-1 fuel cell to complete the uncertainty and sensitivity analysis of effective multiplication factor to resonance self-shielding cross sections. Results indicate effective multiplication factor is the most sensitive to the average number of neutrons released per fission of 235U, and the radiation capture cross section of 238U has the largest contribution to the uncertainty of effective multiplication factor.