An enhanced analytical-based geometry processor for the volume of solid (VoS) method in CPU and GPU computations

Abstract

This paper introduces an analytical–based geometry processor which improves the efficiency of volume-of-solid (VoS) CFD solvers for fluid–structure interaction (FSI) scenarios with rigid body motion. The combination of the VoS method and the direct forcing immersed boundary (DFIB) method speeds up geometry creation while significantly reducing computational workload during unsteady simulations of solid structure movement or deformation. The proposed geometry processor’s effectiveness is demonstrated in the vortex-induced vibrations (VIV) scenario of a single circular cylinder, where the geometry was dynamically modified at each timestep to correspond to the cylinder’s passive motion. Comparisson studies show that the enhanced VoS function accelerates geometry construction process significantly faster than the ray-casting method while maintaining the same level of accuracy. Furthermore, its GPU implementation consistently achieves significant speedup across various computational loads, indicating superior scalability. This proposed method not only works well for simple geometries, but it also supports a wide range of multi-body curvilinear forms, including extruded spans and fully defined 3D geometries, making it suitable for a variety of applications.