Understanding MELCOR CVH/FL Governing Equations for Future AI Applications

Abstract

Severe accidents (SA) in nuclear systems have been studied with reliable system codes such as MELCOR and MAAP. These codes enable the investigation of physical phenomena during SAs by numerically solving thermal-hydraulic governing constitutive equations in a configured geometry with detailed design information. Despite the utility, these system codes have challenges. Configuring the input geometry for packages such as control volume hydrodynamic (CVH), flow path (FL), and cavity (CAV) takes significant amount of time. In order to address the challenge, machine learning (ML) techniques, specifically physics-informed neural networks (PINNs), have been regarded as a possible methodology. PINNs enables physics interpretation by adopting physics equation as a loss function and achieves meshless calculation by incorporating auto-differentiation along with loss function. Due to distinguished characteristics, PINNs is expected to bring simplification to the input for system code, along with the capability of multi-physics analysis for including multiple thermal-hydraulic fields of database. With regards to PINN study, comprehending the governing equation with the logic behind is crucial. In advance to applying PINN to MELCOR, this study aims to assess the feasibility of developing a system code which computes the momentum and the mass of water inside the tank by obtaining the governing equation from MELCOR CVH/FL. Through Python, the forementioned MELCOR governing equation of such flow behavior was assessed by embodying semi-implicit method of calculation.. The accuracy of the developed model was verified by comparison with the MELCOR code in a case study of gravity-driven water tank flow. In conclusion, further application of the governing equation is expected in the future.