We present a novel and scalable approach to accelerate a direct-forcing IBM for moving boundary simulations. Building upon the SIMPLE approach, we keep the pressure and forces coupled, formulate the pressure-forces corrections system, solve for pressure-forces corrections, then perform a projection-correction step for velocities followed by a correction step for pressure and forces. To solve the pressure-forces system, we apply a block reduction and suggest an efficient preconditioning approach that is based on using the Laplacian as a preconditioner for the resulting Schur complement. The Laplacian, being independent of the coordinates of the immersed body, is factorized only once (using an efficient direct method) and is then reused throughout the simulation. We present a rigorous proof of the spectral equivalence of the Laplacian and the Schur complement, thereby showing that the preconditioner's efficiency is independent of the grid resolution. The performance of the method is validated through simulations of the flow around a periodically oscillating sphere in a three-dimensional box. The methodology is then applied to analyze flows generated by arrays of 7 and 14 sub-spheres modeling porous spheres of different porosities. We show good scalability of the algorithm in terms of computational time and memory consumptions. That is, the iteration count for the preconditioned pressure-forces system is nearly constant, regardless of the grid resolution, Reynolds number, and the number of spheres. The developed approach enables accurate moving boundary simulations to be performed on standard workstations, thereby enhancing accessibility of the simulations to the CFD community.
翻译:暂无翻译