Towards Real-time Adaptive Anisotropic Image-to-mesh Conversion for Vascular Flow Simulations
Journal:
arXiv
Published Date:
Dec 16, 2024
Abstract
Presented is a path towards a fast and robust adaptive anisotropic mesh
generation method that is designed to help streamline the discretization of
complex vascular geometries within the Computational Fluid Dynamics (CFD)
modeling process. The proposed method combines multiple software tools into a
single pipeline to provide the following: (1) image-to-mesh conversion which
satisfies quality, fidelity, and smoothness requirements, (2) the generation of
a boundary layer grid over the high fidelity surface, (3) a parallel adaptive
anisotropic meshing procedure which satisfies real-time requirements, and (4)
robustness, which is satisfied by the pipeline's ability to process segmented
images and CAD models. The proposed approach is tested with two brain aneurysm
cases and is shown to satisfy all the aforementioned requirements. The next
steps are to fully parallelize the remaining components of the pipeline to
maximize potential performance and to test its integration within a CFD
vascular flow simulation. Just as the parallel anisotropic adaptation procedure
was tested within aerospace CFD simulations using CAD models, the method is
expected to provide accurate results for CFD vascular flow simulations in
real-time when executed on multicore cc-NUMA architectures.
