Berru-Cms Predictive Modeling Of Nuclear Reactor Physics Systems

Abstract

This Chapter illustrates applications of the BERRU-CMS methodology to paradigm reactor physics systems, starting with the predictive modeling of a paradigm neutron diffusion model of a one-dimensional pool of water containing distributed neutron sources (e.g., spent fuel rods), and continuing with the OECD/NEA (ICSBEP 2010) benchmarks Godiva (a bare uranium sphere), Jezebel-239 and Jezebel-240 (bare plutonium spheres). The paradigm neutron diffusion model characterizes the main physical processes that would occur, for example, in a spent-fuel storage pool and was chosen because it is amenable to closed-form analytical solutions which highlight the salient features of the BERRU-CMS methodology. In addition, this example will also illustrate the basic concepts of computing response sensitivities using the adjoint sensitivity analysis procedure (Cacuci, 1981a, 1981b, 2003). In contradistinction to this didactical example involving the simple neutron diffusion equation, the application of the BERRU-CMS methodology to the benchmarks Godiva, Jezebel-239, and Jezebel-240 involves large-scale neutron transport computations characterized by many imprecisely-known model parameters, as follows: there are 2241 imprecisely known model parameters for Jezebel-239, 1458 imprecisely known parameters for Jezebel-240, and 2916 imprecisely known parameters for Godiva. Eight responses have been considered for Jezebel-239 (the effective multiplication factor, the center core fission rates for ²³³U, ²³⁸U, ²³⁷Np, and ²³⁹Pu, and the center core radiative capture rates for ⁵⁵Mn, ⁹³Nb and ⁶³Cu). Three responses (the effective multiplication factor, the center core fission rates for ²³³U and ²³⁷Np) were selected for Jezebel-240, and eleven responses were selected for Godiva (the reaction rate types listed for Jezebel-239, along with the radiative capture rates for ¹⁰⁷Ag, ¹²⁷I and ⁸¹Br). The advantages of applying the BERRU-CMS methodology in the response-space (as opposed to the parameter-space) for these benchmark systems will become evident, as this methodology will be shown to reduce significantly the computational memory requirements for predictive modeling of systems involving many model parameters. The illustrative examples presented in this Chapter will underscore the fact that the BERRU-CMS-methodology ensures that increasing the amount of consistent information will increase the accuracy of the predicted results, while reducing their accompanying predicted uncertainties (standard deviations).