Prototyping a risk-informed framework to assess safety system effectiveness in pressurized heavy water reactors

Abstract

Simulations play an important role in assessing the effectiveness of reactor safety systems in mitigating accidents. In a pressurized heavy-water reactor (PHWR), the shutoff rods constitute one of two independent safety systems. For Loss-of-Coolant Accident (LOCA) scenarios, the maximum bundle enthalpy is a key Figure of Merit (FoM) in the first few seconds of the accident, as it provides a metric to determine the potential for “energetic” fuel failure. In a risk-informed approach to evaluating the effectiveness of the rod-based shutdown system, the aim is to use statistical methods to show that under all credible initial conditions, there is confidence the shut off rods will be effective in preventing fuel failure. Both epistemic and aleatory uncertainties must be propagated. This can require a very large number of simulations, so surrogate models are needed to reduce the computing time. Our study presents a first attempt in designing such a framework, using a two-phase process: (1) developing a surrogate model to mimic the coupled neutronics/thermalhydraulics LOCA simulations and (2) implementing a two-stage Monte Carlo simulation (TSMC) uncertainty analysis framework to propagate epistemic and aleatory uncertainties. The former is used to provide fast estimates of the simulator code, while the latter produces uncertainty statements about confidence levels for the probability of exceeding a user- defined safety limit. Using the framework, we computed different realizations of the probability distribution functions for the maximum bundle enthalpy predicted by the emulator. We then show how the 99.9th percentile limit values are sensitive to nuclear cross section data. An arbitrary safety limit value is used to illustrate potential applications of the method to safety analyses.