Very Efficient Third-Order CFD Solver
Objective
Develop a third-order method for 3D unstructured grids:
- Very low-cost 3rd-order method
- High-order curved grids not required
- Accurate vorticity (gradient) prediction
- Improved convergence
Motivation
- Development of high-order methods presents a major challenge in large-scale turbulent-flow simulations over realistic complex geometries
- Efficient discretizations and solvers and high-order grid generation remain major pacing items
- Accurate gradients are critical to reliable turbulent simulations, but current methods for unstructured grids are lower-order and prone to numerical noise
Approach
3rd-order edge-based method:
- Quadratic LSQ, flux extrapolation, and source
quadrature make a current CFD code to high-order
Hyperbolic Navier-Stokes method:
- Reformulate the viscous term as a hyperbolic system with gradients as additional unknowns
Progress
- Implemented in NASA’s FUN3D, and accurate gradients demonstrated (AIAA2016-3969)
- Unsteady extension presented at SciTech2017
- Being extended to fully third-order accurate Turbulent flow simulations
