Large Break LOCA Uncertainty Quantification for Boiling Natural Circulation Reactor using Latin Hypercube Sampling

Abstract

Uncertainty is present in every engineering design and it becomes a critical issue in the case of nuclear reactor due to enhance safety concern. For nuclear reactor designs, this uncertainty must be dealt in such a way as to provide ‘acceptable’ confidence that the core will, in reality, operate in a manner that is away from thermal limits than what is predicted from numerical computations.The present work deals with uncertainty in Peak Clad Temperature (PCT) for complete double-ended guillotine rupture of largest diameter pipe carrying cold water in the primary coolant circuit i.e. Large Break Loss of Coolant Accident (LBLOCA) for Indian natural circulation reactor. The double-ended rupture is identified as critical break size leading to maximum clad temperature using best estimate code RELAP5. Six important parameters are selected, which have significant impact of the value of PCT after the initial sensitivity studies.For the uncertainty quantification of PCT, the best method is the application of direct Monte-Carlo technique. However, this is impractical due to very large number of required RELAP5 calculations (~105) and the high cost of each RELAP5 run. To overcome this limitation, Latin Hypercube Sampling (LHS) is used instead of simple random sampling. The LHS technique forces the sampling to select values over the whole range of a model parameter, thereby reducing the total number of samples required to preserve the probability distributions. This significantly improves the computational efficiency of the uncertainty analysis. In the LHS method the range of each input factor is categorized into N equal probability intervals, one observation on each input factor made in each interval.In the current analysis, LHS technique is used to create 500 sets of values of independent variables based on their probability distribution and uncertainty bounds. Further 500 LOCA calculations have been performed using RELAP5 and PCT is obtained for each run. The 95th percentile value of Peak Clad Temperature (PCT) is obtained by directly plotting Cumulative distribution function (CDF). Results are compared with acceptance criteria and found to be satisfactory.