Say I have some mesh nodes that have something wrong with them (aspect ratio, quality, etc.) and they are in an area of importance. How would I go about re-meshing/fixing those nodes? Normally I would delete them, but since they are in an important area I'd have to re-mesh and so far my re-meshing attempts have failed (it re-meshes the entire model over the original mesh and makes for some large file sizes. What would be the best route to take with bad nodes like these that smoothing just won't cure?

Hi,

In my experience deleting and remeshing poor areas does not improve mesh quality. It has never worked for me. I generally use mesh controls and density regions to specify the mesh more carefully and do a new global mesh.

You may also be able to remove the troublesome region into a different body/region and do a hex or prism sweep mesh with better quality and mesh merge the interface to the remaining tet mesh.

Glenn Horrocks

Try moving those nodes manually, if they are not plenty. This option is there in 'Edit mesh' tab. But, this ofcourse needs atmost care.

best luck,

Chirag

I was thinking about creating a density region, but I need to understand how to build a density region and use it a little better. If memory serves there is a section in the tutorials for this and I'll direct my studies there.

As for your second idea, I'm not entirely sure how to do that. How does the mesher move nodes into other regions? Does that node interact with the mesh in that region to some extent? Thanks! -W

Hi,

You can create surfaces which box in a region and give it a different body name. You can now mesh it seperately to the remaining mesh - you could even hex or sweep mesh it. To reconnect it to the remaining mesh you can use the mesh merge function.

Glenn Horrocks

Some good advice there. I've begun to notice a lot of my issues are occuring at corners and nodes around them(no surprise there). I've followed advice from the tutorials but it doesn't seem to help all that much. Hence I've turned my attention to densities and curvature meshing. One problem there of course is that the sizes are refined enough to blow a file size out of proportion very quickly so I'm trying to find the happy medium.

A few models of mine may benefit from the meshing blocks and I'll give it a try when I go to work on them. THanks! -W

Hi,

Just be a little careful with mesh merging. In previous releases it crashed ICEM for me every time, it never worked once except on trivial little test meshes. Any real mesh and it always crashed.

Hopefully now they have resolved the bugs and made it more reliable.

This sort of subdivision is done much more easily and nicely by the unified mesher in workbench. If you want to have some regions swept, some hex meshed and some tet/prism meshed then I would certainly give Workbench a go.

I actually use workbench for the majority of my meshing now as it has caught up on most of the ICEM functionality and it is much easier to use.

Glenn Horrocks

I like the workbench mesher, but I think I posted some time ago that I had problems in defining the internal structures. I've toyed with it to a large extent based on that post but it seems easier to me to go ahead on this particular immersible jet model and use ICEM for the intricate internals. I have found a mesh that looks quite good to the solver, but I'm seeing instability in the residuals which leads me to believe that my turbulence model (SSG) may not enjoy the mesh.

-W

Hi,

For any Reynolds Stress model you have to make the mesh as high quality as possible as these models are very sensitive to mesh quality. I would use high quality hexas as much as possible for a Reynolds Stress Model.

Glenn Horrocks

Hexa models then? What do you think about the body fitted mesh? I've used it with some success but it just seems to me that the body fitted is more structural stress/strain in application.

By the way, I managed to figure out how to mesh out the individual components using CFX-Mesh and it does a fairly good job of creating a quality mesh. Still the same residual problem though.

thanks! -W

Hi,

What do you mean by body fitted mesh?

For RSM models it is worth going the extra mile to generate the best quality mesh you can as they are so sensitive to mesh quality. Therefore if your geometry can be meshed using hexas (you will probably need a ICEM hexa license to do this) then this will probably help. Even high quality tets and prisms can cause problems for RSM models.

I suggest you arrange a trial of ICEM hexa and try to generate some nice hexa meshes. It might just work for you.

Glenn Horrocks

By Body fitted I mean cartesian mesh. In ICEM 11 SP1, they included (or rather finally got to work) a few new mesh types. Body fitted is basically a non-tetra, non-prism mesh that is rectangular in nature. In fact, I absently wonder if this is the hexa mesh we are talking about. When I'm on the ANSYS machine I'll try to find a screenshot and post it.

As for hexa meshing, I can hexa mesh the core of the piece but I don't think that's what you're referring to. I'll call the re-seller to see if there is a license we can look at (and pray that it's cheap). Thanks! -W

Hi,

It is the full structured grid of rectangular elements I am referring to. I don't think hexa-core will help here as it uses tets and prisms to transition from the boundary to the core. The full ICEM-hexa module is what I recommend you try.

Glenn Horrocks

I think we are talking about the same thing. I've sent an email to the re-seller/ANSYS just to make sure but I can create a rectangular grid under the "body-fitted" mesh type. I hope this isn't the only good mesh for the LES/DES/SSG because its a beast to put on my model. I'll see about uploading some pics. thanks! -W

OK, good luck. You will probably find LES and DES are less sensitive to mesh quality so can run on normal meshes. It is just the Reynolds Stress models which are super-sensitive.

Having said that, as LES and DES models require very fine meshes and lots of timesteps, so time spent improving mesh quality initially will be rewarded with a improvements in run time and simulation quality.

Glenn Horrocks

Thanks. I need all the luck I can get.

I've setup a CFX file without the a mesh (its waiting for the final mesh addition for boundary conditions) but basically I'll run the thing for 15 seconds at 0.05,0.25, and 0.5 second intervals (timestep gets larger as things settle out). Also to help, I've created a function which should slowly dial up the velocity and heat. Hopefully this will further prevent any unstable behaviour. Now if I can only iron out those awful negative elements..... =(

Thanks again for the help. -W