I've been running into issues with strong DG on grids which are more complicated than just a periodic cube. To illustrate this as a minimal working example, I've reproduced the issue using a NACA0012 airfoil which is generated using deal.ii (to eliminate any uncertainty about GMSH). Please see a branch on my repository to run these; I have minimally changed the subsonic NACA0012 case to be unsteady and use RK4 such that weak and strong can be compared.
Using weak DG, the result after a small time interval looks reasonable:
But strong DG results in strange jumps, which are aligned with cell faces:
These discontinuities appear at the very first time step. I have seen this on a few GMSH meshes, as has Shruthi, and the commonalities seem to be that we have non-periodic BCs and interfaces between "zones" of the mesh where cell shapes or densities change. The issue seems to go away when grids are perfectly cartesian.
I'm happy to hear any input from others, as this could caused by any number of things, including BCs, the face integral, mesh generation, etc.
I've been running into issues with strong DG on grids which are more complicated than just a periodic cube. To illustrate this as a minimal working example, I've reproduced the issue using a NACA0012 airfoil which is generated using deal.ii (to eliminate any uncertainty about GMSH). Please see a branch on my repository to run these; I have minimally changed the subsonic NACA0012 case to be unsteady and use RK4 such that weak and strong can be compared.
Using weak DG, the result after a small time interval looks reasonable:![image](https://github.com/dougshidong/PHiLiP/assets/94074298/59ff612c-c5a2-4e21-83a1-1665736b2808)
But strong DG results in strange jumps, which are aligned with cell faces:![image](https://github.com/dougshidong/PHiLiP/assets/94074298/31c95a2e-2ab4-44ad-abd4-ec3dd85c3b96)
These discontinuities appear at the very first time step. I have seen this on a few GMSH meshes, as has Shruthi, and the commonalities seem to be that we have non-periodic BCs and interfaces between "zones" of the mesh where cell shapes or densities change. The issue seems to go away when grids are perfectly cartesian.
I'm happy to hear any input from others, as this could caused by any number of things, including BCs, the face integral, mesh generation, etc.