How to get rid of skewed elements while meshing a pipe with many interfacing surfaces
Here is an object that i want to perform analysis on: http://dl.dropbox.com/u/55240438/ima...so_section.png
Here is the geometry i made in design modeler. I had to slice it many times to be able to create 'sweeped method' mesh at those sections. I also made round corners where the fluid makes sharp turns. Then i applied virtual topology onto those corners so that the inflation is not interrupted there.
For some reason i get smooth and even inflation when i use "total thickness" method. Here is what you get: pic1 and pic2
It looks like a good mesh, but it has some bad elements. After playing with the inflation layers number and total thickness size the mesh was improved. After that i tried to raise the "Relevance" value. For some reason its effect cannot be predicted. I got best results at the relevance "20", "21" and "23" and other values just worsened the mesh severely. Also the result on one and the same value is not always the same after regenerating the mesh a few times.
The best parameters i got were
Min Orthgonal Quality: 0,152
Max Skewness: 0.9251
After improving skewness i was able to finally reach convergence in FLUENT but only at k-e turbulence model.
So I wish to improve the mesh more and i need help and advice.
Please, help :)
You may download my project here http://dl.dropbox.com/u/55240438/ima...ipemeshing.rar
I had a quick look at your geometry, first do you have access to ICEM? If yes I should proceed with this meshing software as it allows you to create easily O-grids and permits more "control" than Ansys Meshing.
You can still create O-grid with AM, just split your inner pipe and assign mapped faces (I generate a coarse mesh just to show you the principle). You will have a far better mesh. I would also assign "sweep" methods to your parts, using source and target face in order to "help" the meshing software. I quickly split your geometry and created 2 O-grids. If you spend more time splitting your geometry you could improve your mesh metrics :)
Thank you for the reply!
Well, that looks very good, i didn't know about this method.
But how will the mesh behave at the 'branches' and at the 'turn' when the inflation is added? And as I understand you still cannot avoid tetra cells, and the parts where hexa cells transform to tetra are the most problematic ones.
And, hmm, how did you 'quickly' split the geometry like that? I can think of one way but it's really not quick...))
And I do not have ICEM, unfortunately.
Well for the splitting it took me 10-15min, but once you know how to do it it's quite easy and quick. You just need to split it like I showed it previously. I'm more familiar with ICEM but I use AM for simple geometries, and then use multizone to mesh more "complex" one with ICEM.
For the inflation layers I'm not sure how it will "behave" with AM, I should first assign sweep methods to your parts and then right click on that method and use "inflate this method". And as I said before, for me it's way easier to mesh your geometry with ICEM, but as you don't have it you should continue splitting your geometry into easier "meshable" sub-parts.
If I have some time, I'll try to find a way to mesh your geometry within AM.
Turns out i do have ICEM CFD! That was very pleasant to find out.
So i installed it, but obviously it is going to take some time learning how to use it. I really hope that in the end I would be able to mesh this model with high quality elements. I would be grateful for any kind of further tips.
Thank you, Gwenael, for looking into my model.
By the way, i was able to improve the mesh in Ansys Meshing even a little more just by adding same 'edge sizing' for all of the sweepable elements.The max skewness is now 0.899, min orthogonal quality is 0.159, but... The simple 'element quality' property is as low as 0.043. Do I understand correctly that this property is also important and it shouldn't be that low at all?
minimum orthogonal quality of as low as 0.1 is acceptable
you can also improve your mesh further in fluent by just a few clicks
|All times are GMT -4. The time now is 09:21.|