Efficient Uncertainty Quantification for a Fast-Spectrum Generation IV Reactor System

Abstract

This research is part of on-going research on the management of uncertainties in simulator predictions for Generation IV systems via optimum experimental design. The focus is on uncertainties originating due to input data. The objective is to devise an algorithmic framework for quantification of uncertainties, identification of their key sources, and ultimately guiding the design of validating experiments for their reduction. An integral part of this research is the development of uncertainty quantification algorithms for models involving many input data and output responses. This represents the focus of the research reported here.

Uncertainty Quantification (UQ) in nuclear systems simulation is playing an increasing role in supporting decisions related to the research and development of advanced nuclear energy systems, especially those of interest to the Global Nuclear Energy Partnership (GNEP) and Next Generation Nuclear Plant (NGNP) programs. UQ will help assess the adequacy of existing simulation tools and associated databases, e.g. nuclear cross-section data, and provide guidance to areas of models and/or data where further development and/or measurements should be prioritized. A sensitivity and uncertainty analysis has been conducted to study the effects of neutron microscopic cross-section data uncertainty on macroscopic attributes that influence reactor core design, performance and safety for a Generation IV reactor concept. In the realm of reactor engineering, neutron cross-section data represents the basic physics of neutron interactions with matter and therefore have large impacts on evolution of flux, power, reactivity and other reactor performance attributes. Currently, we focus on uncertainties originating from cross-section data uncertainties, believed to be of primary significance for fast reactor calculations.

This thesis presents a recent development of an UQ algorithm for increasing the efficiency of UQ to a level that enables its execution on a routine basis with best estimate calculations for various reactor performance attributes. Our objective is to devise an algorithm that can characterize uncertainties in the multitudes of reactor performance attributes as evaluated by reactor simulation tools on a routine basis with reference calculations.

The results of this study includes an efficient UQ analysis of calculated uncertainties for chosen key core attributes believed to greatly affect reactor core performance and safety, and an identification of key neutron cross-sections that contribute the most to important core attributes uncertainties.