I am attempting to simulate a turbulent free jet using LES but am having some problems with my mesh.
To create a periodic section of the full 3D geometry I am sweeping a 2-D structured mesh through a wedge. However close to the centre-line the cells become very high aspect ratio and I think this could be problematic.
Below are some pictures of my mesh:
As you can see in the 3rd picture I tried to slice the mesh just above the centre-line and mesh this small area with tet prisms as they capture the geometry better, however I think it is affecting my solution.
My pressure field is clearly affected by the interface between the tet and quad cell regions as can be seen below and I think it is also affecting my temperature field and causing my run to crash:
- Temperature hotspot forms (may be propagating from the strange pressure at inlet/centreline corner)
- velocity skyrockets
- density drops
- enthalpy equation crashes
Below are some pictures of this:
Could anyone please advise me on a good meshing strategy? What is the best way to deal with this tight geometry at the jet centreline?
What software did you use ?? Ansys meshing ?
Slicing is good step, then you have to use multizone method ? did you do that ? alse when you slice, the two bodies have to form one part in designmodeler ?
because of the triangular shape near the center line you need to add like a y-grid...
Hi Ali, thanks for your reply.
I used Ansys meshing however I am importing the mesh into OpenFOAM for solution.
I sliced near the centre-line then swept the top geometries (10 divisions in the swept (circumferential) direction). I then sized the top of the sliced section (10 divisions) so that it matched up.
I am unfamiliar with the multizone method and the y-grid but I will research them and attempt to fix it today.
take a look at this post: http://www.cfd-online.com/Forums/ans...s-meshing.html
if you don't mind sharing your geometry i can try something in icem CFD...
Wow, that would be amazing. I will attach my geometry file (IGES format, I hope that works) and if you could have a go in ICEM that would be great.
In the mean time I will persist with Ansys Meshing :)
Brief description of the geometry:
- LES simulation (just working with really coarse mesh initially ~200k cells to establish stable BC's etc, but will refine mesh considerably later on ~3mil cells)
- converging-diverging nozzle surrounded by free stream
- require named selections for all of the boundary faces (later importing to OpenFOAM
- using wall functions initially at nozzle wall but will eventually solve all the way to wall
As I mentioned earlier, my main concern is with the highly skewed cells near the jet centreline, however I would like to maintain a structured mesh because I'm using LES.
So I had a look at the file you linked me. Can I create an o-grid on the 1/8th section I mentioned, ie would the square cells in the centre just be slightly skewed? I can not afford to mesh the entire geometry (LES will take far too long) so I was hoping to use a periodic section. Would a 1/4 section work better with the right angle at the centreline?
Thankyou Far and Diamondx!
I really appreciate it, I will run my simulation on the new mesh and see how it goes.
Well this looks strangely familar..... Hope your making good progress
I am making progress however I am still having trouble grading my mesh.
Is ICEM in any way like ansys mesher where you can define an edge and a bias ratio etc? Because basically I need refinement at the nozzle wall and nozzle exit and also along the shear layer and inside the jet core.
If anyone has the time and could quickly have a look at it that would be amazing. I attached a link below which includes my project file.
I also attached a picture of how I would like the mesh graded (2D only but I hope you get the idea).
(the project file is in LESnozzle.zip and the picture is called desired_grading)
I also have one more question.
I require the two side faces to be periodic, and I set them as such in the global mesh set-up but I don't think it did anything. When I import into OpenFOAM it says the faces dont quite match up (0.4% error) and it suggests a possible face ordering problem. Is there a way I can establish a cyclic/periodic link between these 2 faces?
If anyone could help me out I would really appreciate it.
defining a bias ration on an edge is one of the easiest thing. in the blocking tab, click on premesh-parametes, then click on edge paramater, on your bottom left you can choose bunching law...
|All times are GMT -4. The time now is 21:49.|