Pore-scale DNS and Upscaling of Turbulent Flows through a Randomly-Packed Porous Medium Using the Method of Volume Averaging

Abstract:

Turbulent flows through randomly-packed porous media are widely encountered in natural systems, for example hyporheic exchange at the interface between free stream water and the underlying sediment, and engineering applications in chemical/nuclear reactors. However, the complexity of the geometry and large scale separation pose huge challenges for detailed examination of these problems using predictive computational approach. In the present work, we first performed direct numerical simulations (DNS) in a randomly-packed triply-periodic porous medium utilizing a fictitious domain method at different Reynolds numbers (10, 50, 100, 200, 300, 600 and 1000). The Eulerian and Lagrangian statistics of turbulence, TKE budget, and anisotropy distribution in confined pore geometries are investigated. Then the closure problem for turbulent flows through packed beds was solved by upscaling the DNS data using the method of volume averaging (Whitaker 1996), where an algebraic model is proposed. Finally, the algebraic model is compared with a conventional, Forchheimer-correction-based model.