Nuclear reactor safety research requires analysis of a broad range of accident scenarios. The major and the final safety defense barrier against nuclear fission products release during severe accident is the containment. Modeling and simulation are essential to identify parameters affecting Containment Thermal Hydraulics (CTH) phenomena. The modeling approaches used in nuclear industry can be classified in two categories: system-level codes and Computational Fluid Dynamics (CFD) codes. System codes are not as capable as CFD of capturing and giving detailed knowledge of the multi-dimensional behavior of CTH phenomena. However, CFD computational cost is high when modeling complex accident scenarios, especially the ones which involve long-time transients. The high expense of traditional CFD is due to the need for computational grid refinement to guarantee that the solutions are grid independent. To mitigate the computational expense, it is proposed to rely on coarse-grid CFD (CG-CFD).

This work presents a method to produce a data-driven surrogate model that predicts the grid-induced local errors. Given the massive high-fidelity data that are produced by either experiments or high-fidelity validated simulations, a surrogate model is trained to predict the grid-induced local errors as a function of coarse-grid features.

The proposed method is applied on a three-dimensional turbulent flow inside a lid-driven cavity. The capability of the method is assessed by applying the trained statistical model on new cases that have different grid size and/or geometry (aspect ratio). The proposed approach is shown to have a good predictive capability.