Comparison of Non-Standard Simulation Methods for Performing Extremely Low Probability Assessments

The probabilistic analysis of possible piping rupture or significant loss of coolant accident (LOCA) is a complex problem as it involves many mechanisms and generates low to extremely low probabilities of events. This topic is of particular interest in the nuclear industry and a conjoint effort between the US NRC and EPRI over the last 10 years has led to the development of the Extremely Low Probability of Rupture (xLPR) code to assess probability of rupture in nuclear piping systems. The current codes for determining the probability of rupture for events occurring at these very low values can take days, or even weeks, to run. In this paper we provide the technical basis for the an adaptive sampling scheme that can reduce the computational time needed for the standard Monte Carlo sampling methods to propagate uncertainty by a factor of 50 to 1,000. A theoretical problem is first presented to explain the method. Afterward the adaptive method is applied to the precursor analysis for time to first leakage.