Speaker
Description
Nuclear data evaluation is a process that takes a long time partly due to resonance identification having to be performed manually by expert nuclear data evaluators. This procedure is laborious, time-consuming, and irreproducible from evaluator to evaluator. Additionally, it is well-documented that a systematic evaluation of the uncertainty on the evaluated cross section is unreliably low. Having these uncertainties that are “too small” results in additional work for the resonance evaluator to search for a non-systematic, artisanal way to make the estimated uncertainty be an appropriate size.
We offer an automated tool for the evaluation process to serve as a starting point for the evaluators by posing the resonance characterization problem as a mixed integer nonlinear program (MINLP). Since the number of resonances present is unknown, the model must be able to be determine the number of parameters to properly characterize the cross section curve as well as calculate the appropriate values for those parameters. Due to the size of the problem and nonconvexity of the parameterization, the optimization formulation is too difficult to solve as whole. We propose a novel method of decomposing the problem into small, overlapping windows that are now solvable, as well as method to stich the windows back together via parameter cardinality and value agreement routines to achieve the global solution. Lastly, a version of quantile regression is run on the windows to provide an uncertainty estimate on the parameter values. The results demonstrate the model being robust to finding the proper number of resonances, the appropriate values of the parameter values, and that the uncertainty estimation is directly reflective of the experimental conditions. Hence, this tool serves as a fast, accurate, and reproducible tool for nuclear data evaluators to have a starting place in their work.