Comparative study on statistical methods for application to best-estimate plus uncertainty accident analysis

Best-estimate plus uncertainty (BEPU) accident analyses can quantify the uncertainties in the calculation results, which will help to reduce the margins in the safety analyses and enhance economics in nuclear power station design and operation while maintaining the safety. The BEPU method requires a generic, reliable and accurate statistical analysis method to determine the upper tolerance limit. The objective of this paper is to compare and choose the statistical method and tool suitable for applications to BEPU. The DAKOTA code was used to perform Monte Carlo sampling on the standard normal distribution function to acquire different samples. Several statistical analysis methods were applied to these samples to determine the tolerance limits, which were compared to determine the most suitable one for BEPU. The analysis results show that the mean value and variance of the upper tolerance limits calculated by the Owen factor method are most close to the analytical solution. When the sample distribution is unknown and there are numbers of input uncertainty variables, the upper tolerance limits can be better calculated by the non-parametric higher order WILKS formula. (authors)