Investigation on Methods for Uncertainty Quantification of Constitutive Models and the Application in BEPU

Abstract

The best estimate plus uncertainty (BEPU) method is recommended by IAEA for nuclear safety analysis. Most of the existing BEPU methodologies rely on the uncertainty propagation of input parameters, while uncertainties of the constitutive models in best estimate codes tend not to be valued or even neglected. A structural method is proposed in this paper to quantify the uncertainties of the constitutive models. Different constitutive models will be classified according to the characteristics and corresponding method could be utilized for each model based on the method. Specific uncertainty quantification (UQ) methods adopted in this paper include the non-parametric curve estimation method, inverse method and design of experiment (DOE) method, and a model selection technique is adopted to opt the optimal model among all alternative models. The structural method is applied to the uncertainty evaluation of LOFT LP-02-6 experiment. Important models are identified, uncertainties of these models are quantified and propagated to the peak cladding temperature (PCT) through code calculations. Uncertainty of the PCT is quantified and the result shows that the calculated values could well envelop the experimental value.