[ICEM] Tips dealing with problem tets/tri manually without complete remesh

FreeFall79 · November 14, 2013, 14:52

Hi helpful community

I am plowing away up this learning curve and have a model going that I am trying to mesh with octatree. I did my refinements and got my settings set and everything looks pretty good except for like a total of 7 elements out of the 9 million in my model. I do not want to wait the 2hrs to mesh again so I tried touching it up manually with the edit mesh tools. My issues of the moment are negative volume, surface orientation and pentrations elements. They are all in the same area so I am hoping I could manually edit. before I tackled my main issue I want to touch up an edge manually for practice. In the screen captures below I did the following

1)Drew a line element between two tri elements in preparation for mesh from edges

2) ran mesh from edges with the close loop and keep volume consistent checked

the problem is it says I have a leak now which I do not get since it seemed simple enough. If figured I would need some best practices for manual edits before I try fixing surface and negative volume tets myself.

Any advice? My only option now is to lower my edge tolerance a bit and pray. There has to be a better way to go about fixing minute issues. I looked at the documentation and searched the forums a bit but could not find anything.

Thanks in advance!

PS IF I am doing a volume mesh do I need to set anything in the global surface mesh parameters to help my meshing issues?

Also is it a good rule of thumb to always surface mesh before volume meshing?

kad · November 18, 2013, 09:56

You can throw away your volume mesh and replace it by delaunay, as the octree mesh transition isn't that good anyway. If there are only seven problematic elements, you can delete them and generate manually tri elements just like you did with line elements. Make sure your surface mesh is "water tight" and has no hole, so run checks. You might also smooth the mesh. After that run the delaunay mesher and check again.

For allmost all geometries the global octree settings are suitable. I change them very rarely. Exceptions are thin gaps. Often errors in octree result from wrong mesh sizes.

FreeFall79 · November 18, 2013, 14:56

Many thanks Kad!!!

I did adjust the mesh size and got a better result. My area and coresponding line mesh sizing were too different i dunnot realy weird since they have a lot of area.

One follow on... Throwing away my octa volume in favor of a delaunay mesh. Do I have to make sure i can run check mesh with no delaunay violations. I have a ton of those which I would have no idea how I would fix.

UPDATE

How can I deal with bad elements on thin areas as the following. I do not want to drop the max element size on these things but I would like ICEM to be smarter about how it meshes in this area. I have tried remesh bad lements but that does not work even with iterations at 100.

kad · November 18, 2013, 19:13

Just give delaunay a try. I never check for delaunay violations before generating the mesh and it works with a probability of 99% as long as the mesh is water tight.

In the thin areas you can use "curve mesh setup" to have a good control over the mesh in these areas. You can preview the node distribution by right clicking curves in the display tree and activate "show node spacing". I think theres no other way than creating a little more elements there. The increase of cell number should not be significant. Remember that insufficient mesh resolution in critical areas can chrash the solver.

You can also refine the surface mesh manually before creating delaunay (its under edit mesh). In the latter and the above case also smooth the surface mesh. The smoother also has the option to refine mesh in areas with bad quality, too. So you see there are lots of possibilties

. Just try what works best for you.

FreeFall79 · November 19, 2013, 17:09

thanks Fad, this is great makes it much easier to see where curve refinement is warranted and does not really impact my mesh times either. I am going to give a this Delaunay, will this make my life easier or harder when I move to place prisms?

FF

kad · November 19, 2013, 18:52

Generally prism should be a little easier with delaunay. The most important thing is to smooth the surface mesh with laplacian smoother before creating delaunay. You can also run alternating cycles with laplacian on and off. This is what I do usually. A very good guide on prism generation can be found in Simon's tips and tricks. It is linked in the sticky thread.

November 18, 2013, 19:13		#4
kad Senior Member Join Date: Feb 2013 Location: Germany Posts: 200 Rep Power: 24	Just give delaunay a try. I never check for delaunay violations before generating the mesh and it works with a probability of 99% as long as the mesh is water tight. In the thin areas you can use "curve mesh setup" to have a good control over the mesh in these areas. You can preview the node distribution by right clicking curves in the display tree and activate "show node spacing". I think theres no other way than creating a little more elements there. The increase of cell number should not be significant. Remember that insufficient mesh resolution in critical areas can chrash the solver. You can also refine the surface mesh manually before creating delaunay (its under edit mesh). In the latter and the above case also smooth the surface mesh. The smoother also has the option to refine mesh in areas with bad quality, too. So you see there are lots of possibilties. Just try what works best for you.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
tips to speedup the convergence of CHT problem with gravity	derisala	FLUENT	1	February 24, 2011 03:28

November 18, 2013, 09:56		#2
kad Senior Member Join Date: Feb 2013 Location: Germany Posts: 200 Rep Power: 24	You can throw away your volume mesh and replace it by delaunay, as the octree mesh transition isn't that good anyway. If there are only seven problematic elements, you can delete them and generate manually tri elements just like you did with line elements. Make sure your surface mesh is "water tight" and has no hole, so run checks. You might also smooth the mesh. After that run the delaunay mesher and check again. For allmost all geometries the global octree settings are suitable. I change them very rarely. Exceptions are thin gaps. Often errors in octree result from wrong mesh sizes.

November 19, 2013, 17:09		#5
FreeFall79 Member Matthew Rich Join Date: Jul 2013 Posts: 33 Rep Power: 13	thanks Fad, this is great makes it much easier to see where curve refinement is warranted and does not really impact my mesh times either. I am going to give a this Delaunay, will this make my life easier or harder when I move to place prisms? FF

November 19, 2013, 18:52		#6
kad Senior Member Join Date: Feb 2013 Location: Germany Posts: 200 Rep Power: 24	Generally prism should be a little easier with delaunay. The most important thing is to smooth the surface mesh with laplacian smoother before creating delaunay. You can also run alternating cycles with laplacian on and off. This is what I do usually. A very good guide on prism generation can be found in Simon's tips and tricks. It is linked in the sticky thread.