Uncertainty Quantification of NEPTUN Analysis Using Wilks’ Formula With an Extended Number of Calculations Available to Purchase

The two-phase system analysis code MARS [1] has been used for the uncertainty quantification of NEPTUN reflood test [2] analysis. By performing 10,000 calculations based on a random variation of the MARS model parameters and measured data, a mean value, and the 95% upper bounds are traced along the number of calculations. The CPU-intensive calculations were performed using the 11 node PC-cluster under Linux platform. The Monte-Carlo calculation results suggest a total number of 2,000 calculations would be sufficient to determine the stable mean and 95% upper bound values. The peak temperatures predictions are also used to find the 95% bounding values by using the Wilks’ method. For the 1^st order one-sided formula, every 59 peak temperatures are examined to locate the bounding temperature, with a 95% confidence. The 2^nd and 3^rd order values were found in a similar way. The uncertainty band by the Wilks’ formula, when compared with the true 95% bounding value, is observed to be quite broad, especially in the case of the 1^st order. The 2^nd or 3^rd orders or a full Monte-Carlo method would be necessary to demonstrate that the safety of the plant is ensured with a sufficient margin. A supplementary sensitivity study, for the nine uncertain parameters selected for the NEPTUN analysis, is also performed to find the degree of influence of each parameter on the peak rod temperature.