Uncertainty Quantification of Physical Models and Extrapolation of Uncertainties during LBLOCA

摘要

The use of the best-estimate (BE) computer codes in safety analysis for large break loss-of-coolant accident (LBLOCA) is the major trend in many countries to reduce the significant conservatism. A key feature of this BE evaluation requires the licensee to quantify the uncertainty of the calculations. Uncertainty includes those of physical model and correlation, plant operational parameters, and so forth. The statistical methods in uncertainty quantification process are used to reasonably determine the uncertainty ranges of physical models instead of expert judgement. Initially, the most influential parameters are determined based on the sensitivity analysis on each parameter or ranking of Pearson 's coefficient on each parameter obtained by randomly sampled calculation with the preliminary fixed uncertainty ranges. Those analysis are performed with respect to the twenty-one influential parameters associated with reflooding phenomena (highly ranked parameters from Rod Bundle Heat Transfer program performed by USNRC). In this study, CIRCE method and MCDA method are used to quantify the distribution of the most influential parameters during reflood phase of LOCA in MARS-KS thermal-hydraulic code. CIRCE method, developed by CEA in FRANCE, is an inverse method using the E-M (Expectation-Maximization) algorithm based on the principle of maximum of likelihood and Bayes' theorem. MCDA method has been developed by KAER1 in Korea based on Bayesian statistics as well. FEBA reflooding tests are chosen to quantify uncertainty ranges in the CIRCE and MCDA methods. With the determined uncertainty ranges in two methods, envelope calculation with randomly sampled uncertainty parameters for FEBA tests themselves are performed to check and confirm whether the experimental responses such as the cladding temperature are inside the limits of calculated uncertainty bounds. Recently, one of the main questions raised in uncertainty quantification, is the capacity of extrapolating the quantified uncertainties of input parameters which is determined in the small scale separate effect test (SET) experiments to the larger scale experiments or nuclear power plants. For those application, more uncertainty parameters or different set of parameters may be needed because of the different thermal-hydraulic behavior or multi-dimensional effects. So, a confirmation step through uncertainty evaluation is conducted to estimate the contribution of the uncertainty parameters in the case of the large scale integrated effect test (IET) experiment like LOFT L2-5 and the commercial nuclear power plant during LBLOCA.