Fluent diverges when using inflation layers
I'm simulating a 2D wedge in a flow with the transient standard k-epsilon model. The Reynolds number is around 400000.
The geometry is shown below. The mesh without inflation layers gives a solution with a CD of around 1.2. This is close to literature values of a 2D wedge in a flow. However, when I add the inflation layers (and change nothing else in the mesh), the solution diverges after one timestep and I get a warning that the turbulent viscosity is limited to a viscosity ratio of 1e5 in several thousand cells. In the end Fluent stops iterating with the message: Code:
Divergence detected in AMG solver:y-momentum I've read up on the viscosity ratio warning in different posts here and in other internet resources, but the remedies (refining the mesh, under-relaxing the k and e) did not work for me. I'm surprised that adding an inflation layer has such a big (negative) effect on my solution. Can someone explain what is happening here? Should I just leave the inflation layers out all together (sounds like a bad idea)? 1. Geometry overview, flow comes from the left, normal to the boundary: https://dl.dropboxusercontent.com/u/390388/geo.png 2. Geometry close-up: https://dl.dropboxusercontent.com/u/.../geo-close.png The mesh without inflation layers looks as follows: 3. Mesh overview: https://dl.dropboxusercontent.com/u/390388/mesh.png 4. Mesh refined area: https://dl.dropboxusercontent.com/u/...esh-close1.png 5. Mesh around wedge: https://dl.dropboxusercontent.com/u/...esh-close2.png And with inflation layers: 6. Mesh around wedge with inflation layers: https://dl.dropboxusercontent.com/u/.../inflation.png 7. Mesh around wedge nose with inflation layers: https://dl.dropboxusercontent.com/u/...tion-close.png |
Have a look at the cells at the sharp trailing edges of your geometry.
I guess they have very small angles. |
The trailing edge looks like this:
https://dl.dropboxusercontent.com/u/390388/te.png How do I prevent too sharp angles in the mesh? |
An easy way could be to refine the mesh in this region (which you should do anyway because the mesh is rather coarse here).
|
I've refined the mesh at the trailing edge by adding an edge sizing restriction on the edges of the wedge. It now looks like this:
https://dl.dropboxusercontent.com/u/...ion-refine.png Fluent does converge now. Thank you for your help. |
All times are GMT -4. The time now is 01:49. |