Regenerating and estimating missing parameters in critical benchmark experiments: A framework for inverse uncertainty quantification using information-theoretic experiment design

Abstract

The verified benchmark data are integral validation components of contemporary neutronics calculation, which plays a crucial role in advanced modeling and simulation, criticality and reactor safety analysis, and nuclear data testing. Nevertheless, the phenomenon of retained experiment data loss has occurred occasionally in recent years, causing great difficulty and influence on the formulation and compilation of benchmark in international evaluation projects. Therefore, we proposed a comprehensive approach to address the issue of data missing in benchmark experiments. In this paper, we use the Bayesian inference framework, combined with the information theory method, to give a general process framework for dealing with missing data estimation. This framework spans from experimental strategies design to experimental data acquisition, and further to Bayesian inverse uncertainty quantification (IUQ) based on the obtained data. It indicates that this approach can effectively reduce the uncertainty of unknown parameters with fewer IUQ evaluation times, and provide the effective estimation of missing data. The results obtained her e not only provide a theoretical basis for regeneration and estimation methods of missing parameters in critical benchmark experiments, but also have great engineering application prospects in parameter estimation and model calibration.