Research on impact of cross section consistence and different covariance library in the process of uncertainty analysis

[Background] The nuclear data is the fundamental input parameter of reactor physics computation and it is also one of the significant uncertainty sources to the result of nuclear physics calculation. Usually, the cross section is measured by the integral experiment. However, the cross section obtained from the experiment is always imprecise due to numerical approximation and model simply in the experiment. [Purpose] This study aims to evaluate the effect caused by relative factors of cross section consistence and different covariance library in the process of uncertainty analysis. [Methods] The home-developed code SUACL based on statistical sampling method was developed for the uncertainty analysis, and the Three Mile Island unit 1 (TMI-1) benchmark was taken to evaluate the impact of both the initial value of cross section and the various covariance matrices. [Results] It is found that the larger of the cross section value, the higher uncertainty caused by it in the same degree of perturbation. Therefore, it is better to use the practical initial value of cross section. The uncertainty of infinite effective multiplication factor would be different based on different evaluated neutron data library such as ENDF/B-VII.1, JENDL4.0 and JEFF3.2. For example, the uncertainty of ²³⁸U fission cross section based on JEFF3.2 is smaller four orders of magnitude than that of ENDF/B-VII.1 and JENDL4.0. [Conclusion] The appropriate covariance data and precise unperturbed initial value of cross section are critical to analysis uncertainty. Both of them have great impact on the uncertainty result. (authors)