Automated Voronoi Meshing for High-Fidelity Aerodynamic Simulations
Recent advancements in high-performance computing have accelerated the adoption of wall-modeled large-eddy simulations (WMLES) as a tool for practical applications in the aerospace industry, since WMLES can considerably increase the prediction accuracy of simulations of unsteady flow phenomena at a reasonable cost. However, one bottleneck that prevents a more widespread usage of WMLES in flow simulations around complex geometries is the challenge of creating the large-scale, high-quality meshes required. To tackle this issue, computational fluid dynamics experts at NASA’s Ames Research Center are currently developing an unstructured solver-mesher pair tool for the agency’s Launch, Ascent, and Vehicle Aerodynamics (LAVA) software. While traditional approaches for mesh generation around an airplane in a high-lift (takeoff or landing) configuration can take a handful of specialists multiple weeks to complete, unstructured Voronoi grids reduce the computational and human resources needed to generate a high-quality grid to just one person over a couple of days. These grids are made up of clipped convex polyhedra and offer highly automated and scalable body-conforming grid generation. This new tool in LAVA has already been used across a number of problems of engineering relevance—from analyzing the aerodynamics of commercial airplane configurations to simulating the sound pressure levels generated by supersonic jets with various nozzle geometries.
Quick Facts
The development of the unstructured mesher-solver is certainly a leap forward in simulation capability. First, it enables the generation of meshes around arbitrarily complex geometries at a fraction of the cost used by traditional methods. Second, it allows for consistent grid refinement, grid convergence studies and, ultimately, an increased reliability on simulated results.
- The LAVA unstructured solver was able resolve differences in cells that were four millimeters apart in the flow field surrounding half of a 60-meter-wingspan aircraft in a high-lift configuration. The simulations ran over two days and required 100 AMD “Rome” nodes of NASA’s Aitken supercomputer.
- The team found that the LAVA unstructured Voronoi mesher-solver tool greatly reduced the amount of hands-on work necessary to generate a mesh around complex geometries, while producing accurate results on par with hand-tailored curvilinear overset meshes.
- Development of an automated mesher and efficient solver capable of rapidly conducting WMLES is important to enable NASA’s aeronautics goal to have the capability to certify aircraft readiness through the analysis of simulations, in addition to helping engineers decrease design uncertainties of new vehicles and configurations, such as the noise generated by future supersonic commercial jets.
- Currently, the LAVA team is working on increasing the performance of the new mesher by introducing a load-balanced parallel computation of the unstructured Voronoi mesh, as well as improving the solver by using GPUs to reduce the time each simulation needs to run.
Researchers
- Victor Sousa, NASA Ames Research Center Please enable Javascript to see email address.
- Emre Sozer, NASA Ames Research Center Please enable Javascript to see email address.
