Large Eddy Simulation of Flow Interactions Between a Turbulent Free-Stream and a Permeable Bed
Benjamin H. Sonin
DOI: https://doi.org/10.48550/arXiv.1707.09541
IF: 0.673
2017-07-29
Fluid Dynamics
Abstract:Systems comprising a turbulent channel flow overlaying a permeable bed can be found in a variety of industrial and natural applications (e.g. urban planning, fracking, submerged vegetation). One important realization of this system is at the bottom of rivers, where surface waters of the river permeate the deposited sediment and exchange their contents with subsurface waters. Obtaining a complete picture of this process, known as hyporheic exchange, is important for the understanding and maintenance of water quality and river ecology. Given the wide range of length scales (mm to km) and corresponding time scales of interest, reduced-order numerical models (e.g. stochastic particle tracking, advection-dispersion equations) are often used to study the transport of mass and momentum in this system. However, the many scales associated with surface/subsurface mixing make the development of a high-fidelity model for turbulent scalar transport a formidable challenge. In an effort to capture transport behavior at the pore scale, this work is focused on development of a detailed, three-dimensional flow model for the Large Eddy Simulation (LES) of turbulent flow over a permeable bed. A rigorous assessment of the computational mesh, domain size and turbulence model is performed. Concurrently, the double-averaging methodology is used to study the large-scale, persistent flow behavior in the presence of spatio-temporal heterogeneity. It is found that accurate resolution of the largest turbulent scales is necessary to develop a complete picture of interfacial transport, and that these large structures are felt deeper within the bed. Additionally, the development of a particle tracking software module is detailed. This module provides a foundation for the extraction of Lagrangian dispersion information from the LES, which will eventually be used to physically inform reduced-order models.