Development of a Nonintrusive ROM for Parametric Study in the RPV Lower Plenum

Abstract System thermal hydraulic codes are extensively used to assess the response of large-scale systems due to their computational efficiency in simulation. Nevertheless, they cannot simulate inherently multi-dimensional physical phenomena because of the one-dimensional approach adopted. A common example is the multi-dimensional flow in the reactor pressure vessel lower plenum. One solution is to employ computational fluid dynamics codes to simulate the lower plenum and obtain the required multi-dimensional flow. However, computational fluid dynamics computation is typically time-consuming, introducing a new challenge. In this paper, a nonintrusive reduced order model for steady state calculation of two-dimensional lower plenum has been developed using proper orthogonal decomposition and sparse grids methods to overcome this challenge. The input of reduced order model is the coolant mass flow rate at the inlet of the lower plenum, and the output is the velocity field distribution in the lower plenum. The range of the inlet flow rate is from 500 kg/s to 600 kg/s, and all data in snapshots are generated through repetitive simulation under different sampling points using OpenFOAM. The performance of the constructed reduced order model is validated on a test set including 16 new sampling points. The results indicate that the established reduced order model can rapidly provide velocity field distributions in the lower plenum for any inlet flow rate ranging from 500 kg/s to 600 kg/s, and the errors are all acceptable to meet simulation requirement.