Efficient simulation time reduction in uncertainty quantification via the polynomial chaos expansion method

Abstract

In this paper, the use of the polynomial chaos expansion (PCE) method in uncertainty quantification (UQ) is adapted in two-step method STREAM/RAST-K to enhance the simulation efficiency and minimize the memory consumption. In neutronics analysis filed, traditional stochastic sampling methods require significant computational resources for simulations with perturbed parameters, resulting in extensive memory usage and prolonged simulation times. Furthermore, the two-step calculation process further extends the simulation time, particularly during transport calculations, owing to the requisite branch calculations for generating few-group constants. To address these challenges, PCE was incorporated for UQ regarding the perturbation of input parameters, such as UO(2 )enrichment, UO2 density, pellet radius, and clad outer radius. The uncertainty analysis in modeling benchmark was employed for the calculations. Both square- and hexagonal-structured fuel assemblies were analyzed using the TMI-1 and Kozloduy-6 benchmarks. Regarding the two-step method, UQ with PCE was found to be at least 20 times faster than UQ without PCE over a burn-up range of 0-30 MWd/kg, with the uncertainty differences being negligible (<0.005 %). The results and findings of this study effectively demonstrate the benefits of integrating PCE into UQ, notably in decreasing the simulation time by limiting the number of perturbed samples during the UQ process.