The Coupling Scheme of FVM and GPU-accelerated LBM for Complex Geometries
Please login to view abstract download link
The GPU-accelerated lattice Boltzmann method (LBM) is considered a potentially valuable tool for high-fidelity simulations, which could be more efficient than traditional CFD methods running on central processing units (CPUs) [1]. However, when the flow over a complex terrain is considered, an LBM solver could suffer from the need for very fine grid resolutions near the wall region, thus involving a higher computational cost. In this work, we assess the potential for a coupled numerical procedure, where near-wall quantities from an underlying RANS model computed with a finite-volume solver running on CPU are coupled to an LBM solver running on GPU [2]. The coupling scheme is validated against a turbulent channel flow with a friction Reynolds number of 180. A sensitivity study on the coupling parameters is also carried out, and the results demonstrate good agreement with data from direct numerical simulation. Ongoing work involves the validation of the coupled approach for complex geometries, particularly examining the interaction between the atmospheric boundary layer and a two-dimensional hill simulation.
