A non-intrusive reduced order model for light water reactor core stability analysis

A non-intrusive reduced order model for studying the nuclear reactor core stability in presence of localised perturbations due to operational uncertainties is analysed. Following the standard neutronics calculation chain, a two-step model based on a combination of POD and RBF approaches is proposed in order to reduce the computational time of both the cell calculation, devoted to produce homogenised data for the full-core diffusion calculation, and the diffusion calculation itself. The results presented suggest that, while the cell model seems adequate to reproduce untrained operating conditions, the diffusion model still needs to be properly tuned, improving the training set quality.