Confidence Improvement Effort
Baptiste Mouginot
Abstract
As fuel cycle simulation tools aims to simulate prospective nuclear fuel cycles, and because of the sensitivity of the actual nuclear fuel cycle data, it is extremely difficult to establish confidence in any single tool without a new major code comparison project. Uncertainties associated with fuel cycle modeling such as costs, time-frames, and physics of alternative technologies often make the validation of simulation results complicated. Classical code comparisons and benchmarks have generally suffered from an inability to compare results between simulators due to differences in modeling choices and output data [1-2]. Currently, different institutions are developing their own numerical tools regarding their scientific objective. These tools may have several different philosophies and thus cannot answer all the relevant issues of future nuclear fuel cycle. One partial solution, to resolve this shortcoming within the community, will be to build a general and universal framework dedicated to the validation of simulation tools. Within this framework, we a better understanding of the main sources of uncertainty and comparison obstacles could be achieved, There are three main sources of uncertainty that affect most fuel cycle simulations:Approximation and hypothesis modeling non-existent reactors/fuels.Fundamental data uncertainties that propagate through the fuel cycle simulation to the output.A wide range of input variables and parameters are possible, and this can strongly impact the result of a scenario simulation.While points 1. and 2. are already being investigated by fuel cycle tool developers in other research frameworks, this work aims to address point 3.Using sensitivity analysis (SA) formalism will help in both understanding and improving the weight of input variables and parameters via evaluating output standard deviations. In the framework of this open collaboration, we aim to build a set of experiments composed by a set of subtasks of growing complexity. This allows each component of a fuel cycle tool to be evaluated in depth. The experiments will be concluded with an SA on the full problem, yield a deeper understanding of the constraints and limits of a specific tool. The ability within the framework to compare results with other fuel cycle tool calculations will provide clearer requirements of certain tool-specific features to explain/understand some behaviors. This first experiment has been designed as a simple test case to assess the correlation of a once-through cycle composed of an enrichment facility, a PWR loaded with UOX fuel, and a storage facility for used UOX fuel. Both CLASS[3] results and Cyclus[4] results and their comparison will be presented.