SDF-Guided Point Cloud Generation Framework for Mesh-Free CFD

Abstract

Meshing is a bottleneck of CFD workflows, especially when complex geometries are considered. Mesh-free methods could be a promising solution, but the lack of high-quality point cloud generation methods for boundary layers has hindered their popularity and applications. This work presents a novel point cloud generation framework for near- and off-body regions. The novelty of the method is the introduction of the Signed Distance Function (SDF) to guide advancing point layers in the near-body region. Insertion/removal mechanisms of points, collocation search approach, and point cloud quality metrics were also proposed. These ensure high-quality boundary layer resolution in the near-body region, regardless of the complexity and topology of the geometry. For the off-body region, Cartesian points are employed for smooth and adaptive point distributions. Compared to conventional advancing front point generation, the proposed method ensures surface-norm point distributions with consistent layer structures, which are critical for boundary layer resolution. Compared to the strand mesh generation, the current method presents much greater flexibility with few restrictions on inter-layer connections. The proposed approach is tested for various 2D and 3D benchmark geometries, along with mesh-free modeling results using the generated point clouds. The results demonstrate an important step towards a fully automated, adaptive, and mesh-free CFD workflow for complex engineering applications.