Systematic Sensitivity and Uncertainty Analysis of Sodium-Cooled Fast Reactor Systems

Sodium-cooled fast reactor (SFR) technologies have the potential to guarantee energy supply and to reduce the burden of nuclear waste for future generations. For an adequate simulation of these reactor systems, well-established tools that have so far been applied mainly to light water reactor (LWR) concepts need to be validated and enhanced.

For licensing purposes, there is an increasing interest in replacing conservative calculations by best-estimate calculations supplemented by uncertainty analyses. Nuclear data are a major source of uncertainties in reactor physics calculations. The propagation of nuclear data uncertainties to important system responses is important for determining appropriate safety margins in safety analyses.

A systematic approach for quantifying nuclear data--induced uncertainties for all stages of modeling is needed to assess the performance of traditional methods for uncertainty and sensitivity analysis and to unveil the major drivers of observed uncertainties in SFRs. This thesis presents a basis for such a systematic approach through the use of sub-exercises that address different levels of modeling as addition to the OECD/NEA Benchmark for Uncertainty Analysis in Modelling of SFRs.

The major analysis method applied within this thesis was the random sampling--based XSUSA method in which nuclear data is varied based on the corresponding covariance data. As a basis for analyses using several multigroup neutron transport codes from the SCALE code system, new multigroup cross section and covariance libraries were developed and optimized for the analysis of SFR systems. In order to use the time-efficient XSUSA method in combination with the SCALE 6.2 release, SCALE's random sampling sequence Sampler was extended to allow the perturbation of cross sections after the self-shielding calculation, including an optional approximation for consideration of implicit effects.

XSUSA allowed for the determination of one correlation-based sensitivity index to identify the main contributors to observed uncertainties. This sensitivity analysis was extended by a second correlation-based sensitivity index, as well as variance-based Sobol' sensitivity indices. Furthermore, corresponding indices that use sensitivity coefficients from perturbation theory were developed to allow for comparisons between the various approaches.

Finally, systematic uncertainty and sensitivity analyses with respect to nuclear data were performed based on the developed specifications and the described developments. It was found that the analysis of simple models is sufficient for initial assessments of the impact of nuclear data uncertainties on larger scale models as well as the corresponding identification of the uncertainties' major drivers. In general, significantly larger uncertainties for eigenvalues and reactivity coefficients were observed than in corresponding LWR calculations. The main contributor to the uncertainty for most output quantities was iden