How to avoid single edges at boundaries in 2D models?
 

Hello everybody. I am new in ICEM CFD meshing. I have done some tutorials about hexa meshing, and I now I am trying to simulate in Fluent the movement of a fluid in a kind of ring by natural convection (like a thermosiphon).

I have made a 2D model and I have meshed it, by blocking strategy, pre-meshing and conversion to unstructured mesh. The quality of the resulting mesh is good, but I have found a big problem: when I check the mesh to find errors and possible problems, single edges reveal in all the elements of the domain boundary.

After this, I have tried meshing simple 2D shapes, as a square. I have created the geometry (4 points, 4 curves and a surface) and I have meshed by using different methods (blocking; surface meshing by Mesh>Compute mesh>Surface mesh only) but the problem of single edges at boundary always returns.

How can I avoid it? Thank you very much in advance!

attach tin and blocking file

Far, the files are now attached. Thanks.

hi
 

Hi

you don't need to worry about the single edges this isn't a problem.

you can continue with the analysis.

Thanks
Kiran

Thanks, Kiran.

Today I check one of my mesh on 32 and 64 bit systems. On 64 bit system I dont get this problem whereas on32 bit I am getting this problem. I guess this is due to round off error since meshing spacing is smaller than computer round off error.

When opened your files, I am not getting any error or warning message. Strange !!!!

Far, I have imported the mesh to Fluent, as Kiran said, and solved a case. Non error or warning message has been shown, and the solution seems valid.

Thank you both.

For a little more explanation...

Single edges is on the list of "Possible Problems". If you were doing a 3D analysis and didn't have any baffles, then single edges would be a concern.

But you should expect single edges along all your boundaries for 2D analysis. You should also expect single edges along the edges of baffles in 3D, etc.

What does single edges imply in the mesh? Like we know that skewness shows the distorted cell in the mesh.

Single - double - multiple...
 

A double edge of a shell element is an edge shared by two shells (normal situation). A triple edge is an edge shared by three shells (like a T-connection), which could be normal if you expected a T-connection or could suggest a problem. Rather than have a special diagnosis for "triple edge", we just have one for "multiple edge", which could be three or more shells meeting at an edge...

By the same logic, a single edge is an edge connected to only a single shell... In practice, it is the perimeter of a 2D or zero thickness shell model. Perfectly acceptable in those circumstances, but potentially a problem if you were trying to form a closed volume so you could fill with delaunay or hexa dominant fills...

When the mesh check tells me I have one of these situations, even if I am expecting it, I just ask for the subset. Then I add one or two layers to the subset so I can get an idea of what is going on. You can also right click on shells to have it highlight these single or multiple edges. Either way, don't decide if it is a problem or what to do until you have seen it in context.

Hi, Simon!!

Just a simple question... I am trying to mesh a 2D circular cylinder, so you can imagine which my problem is: single edes! also you told single edges is not always a problem (above all in 2D meshes), if I export my bidimensional mesh to fluent without checking, I found an error from fluent when reading. So, could you tell me how to face this problem?? I am sure it has to do a very easy solution as I was getting excellent results from 3D hexa meshes...

Best,
Antonio

Hello everybody,
i am using icem cfd for 1 year and i am creating my first 3D mesh.
My cfd calculation give me an error with that mesh only on some elements.
Mesh quality is good for icem but there is a problem with some single edges,
these edges and faces are precisely the elements making problems during the
computation.
I dont' know how to deal with that problem, would you know what to do ?
Thanks
Jérôme
http://hpics.li/9077c88