Maximizing accuracy in severe accident simulations: An In-depth analysis of sampling methods for MELCOR code during station black-out in WWER-1000

This study explores the uncertainty in severe accident analysis of nuclear power plants, focusing on the thermal–hydraulic parameters of the WWER1000 reactor during station black-out scenarios. The primary challenge in such analyses lies in the inherent uncertainties associated with the simplifying assumptions embedded in computational codes. To address this, five prominent sampling methods—simple generation sampling, Latin hypercube sampling, SOBOL, HALTON, and LHS-SOBOL—are applied to improve the MELCOR code’s predictive accuracy. The results highlight that SGS produces non-uniform input distributions and performs poorly in comparison to the other methods. Additionally, key parameters, including hydrogen production, maximum containment pressure, and reactor pressure vessel rupture time, show significant variability, ranging from 150 to 602 kg, 14,100 to 28,000 s, and 0.62 to 1.05 MPa, respectively, with a 95 % confidence level. The findings emphasize the importance of selecting robust sampling techniques for enhancing the reliability of nuclear safety assessments.