Review of Physics-based and Data-driven Multiscale Simulation Methods for Computational Fluid Dynamics and Nuclear Thermal Hydraulics

Abstract: Modeling of fluid flows requires corresponding adequate and effective approaches that would account for multiscale nature of the considered physics. Despite the tremendous growth of computational power in the past decades, modeling of fluid flows at engineering and system scales with a direct resolution of all scales is still infeasibly computationally expensive. As a result, several different physics-based methodologies were historically suggested in an attempt to "bridge" the existing scaling gaps. In this paper, the origin of the scaling gaps in computational fluid dynamics and thermal hydraulics (with an emphasis on engineering scale applications) is discussed. The discussion is supplemented by a review, classification, and discussion of the physics-based multiscale modeling approaches. The classification is based on the existing in literature ones and distinguishes serial and concurrent approaches (based on the information flow between micro- and macro-models) with their variations. Possible applications of the data-driven (machine learning and statistics-based) methods for enabling and / or improving multiscale modeling (bridging the scaling gaps) are reviewed. The introduced classification distinguishes error correction (hybrid) models; closure modeling (turbulence models, two-phase and thermal fluid closures); and approaches aimed at the facilitation of the concurrent physics-based multiscale modeling approaches. A comprehensive review of data-driven approaches for turbulence modeling is performed.