Basic questions about workflow, grids, viewing

[bbbeard]

Hi, I am still very much a newbie with Elmer, though I have some experience with other FEM solvers. I am using Elmer on a Windows Vista64 Thinkpad T61 platform and am also using it on a Windows 7 Starter(!) netbook platform.

I am confused about some basic concepts.

(1) It seems that only a handful of solvers are exposed in the ElmerGUI Model | Equation... menu. I gather that many more solvers are available because the share/elmersolver/lib directory has a bunch of enticingly named DLLs. I can't find where this is explained but it seems that the ElmerTutorials.pdf file exhibits many of these solvers. The typical chapter in ElmerTutorials.pdf explains the problem, shows the structure of the *.sif file, and then shows the results. Is the basic workflow: (a) find some way to generate a grid (ElmerGrid, Netgen, Tetgen?), (b) Compose a sif file, (c) open Elmergui, (d) load these files, (e) Click the button labeled "S", (f) wait for the Solver to finish, (g) view in ElmerPost or otherwise crunch these results? How does one load a sif file? I see there is a menu item File | Load Mesh... but not a File | Load SIF.

(2) I was able to do some nice simple FEM work using Netgen to create a Delaunay tetra mesh and then ElmerGUI linear elastic solver and then R for the post-processing (I needed to dig a little deeper in the stress distribution than the high-level ElmerPost results). However, whenever the mesh contained more than about half a million elements, Elmer would lock up. But elsewhere I read about Elmer being used for problems with billions of nodes. So what is the direction I should push? Is there a magic setting that allocates more memory for Elmer? Should I try to set up the parallel processing capability?

(3) I had a great deal of trouble selecting the surfaces to apply boundary conditions to. It seemed that whenever I clicked on a surface I grabbed the surface behind it, or rather, whichever was the outer surface. Specifically, the geometry I was dealing with was a large cylinder with a smaller cylinder removed from it (in the sense of CSG subtraction) so as to create a hole in the bottom of the larger cylinder (like a wine bottle punt but cylindrical). I had no trouble grabbing the top and the sides and the annulus on the bottom of the large cylinder, but I was never successful grabbing the inner surfaces (formed by the smaller cylinder). What am I missing? Aggravating this problem is the fact that I can't figure out how to zoom the view of the part in ElmerGUI. I can rotate it with the left mouse button, but the right button brings up a menu. What's the trick? I am doing this on a Thinkpad with a Synaptics Touchpad and the Thinkpad pointing stick. Also, because I couldn't zoom, using the option of showing the surface number for each node is a loser -- the numbers are unreadable for the finer meshes!

(4) I need to get to the point where I can do a 3d fluid potential flow analysis with a complex-shaped moving solid surface (a "header") and a free liquid surface. I really just need the kinetic energy distribution as a function of position for a unit velocity for the header. But the mesh needs to be fine enough to capture the streamlines around the complex-shaped header. I can't seem to get Netgen to put enough nodes around the header, even with the mesh set to "very fine". ElmerGrid doesn't seem to handle complex geometries very well. Any suggestions?

Thank you for this excellent tool, by the way! It is really important to my work to be able to post-process the output files precisely, node-by-node, and that is one of your strengths.

[raback]

Hi

1) Many questions are answered in the FAQ:
http://www.elmerfem.org/wiki/index.php/ ... _Questions

2) To save memory try going from Direct to iterative methods, and expensive preconditioners to cheaper ones. Of course this does not often work due to convergence problems. For power users the solution is often to use parallel computation. See the dedicated chapter in the ElmerSolver manual on that.

3) Normally the zoom is available on the center roll in the mouse. If that's not available I think that pressing the both mouse buttons together could do the trick....

4) Many use GiD which is not free but it is cheaper than most commercial alternatives.

If you need detailed postprocessing features you might have a look what SaveLine and SaveScalars routines have to offer. Also you could use ResultOutputSolver to save data in .vtu format for further work with Paraview, for example.

-Peter

[petroo]

Hello bbbeard,

in addition to "Peter the Guru's" remarks, let me add some smaller hints that I don't feel covered by the other Peters answer. Maybe they help.

(3) I had a great deal of trouble selecting the surfaces to apply boundary conditions to. It seemed that whenever I clicked on a surface I grabbed the surface behind it, or rather, whichever was the outer surface. Specifically, the geometry I was dealing with was a large cylinder with a smaller cylinder removed from it (in the sense of CSG subtraction) so as to create a hole in the bottom of the larger cylinder (like a wine bottle punt but cylindrical). I had no trouble grabbing the top and the sides and the annulus on the bottom of the large cylinder, but I was never successful grabbing the inner surfaces (formed by the smaller cylinder).

Maybe you missed the function to hide a front surface? If you try to select a certain surface but end up with one in front of it, push Ctrl-H (for "hide surface") to get rid of it w/r/to selection. After that you may point-and-click to the formerly hidden, back-staged ones.

Aggravating this problem is the fact that I can't figure out how to zoom the view of the part in ElmerGUI. I can rotate it with the left mouse button, but the right button brings up a menu. What's the trick?

Peter suggested the joint pressing of the 1st and 2nd mouse button. I fear that this will not do the job, as this combination usually emulates the 3rd (middle) button of the mouse device. The scroll wheel usually is interpreted as a 4th and 5th "button", thoughm, IIRC. Why not get a simple USB mouse with a scroll wheel and attach it to the Thinkpad?

ElmerGrid doesn't seem to handle complex geometries very well. Any suggestions?

For my work, Salomé (OS system) proved the most effective system to create reasonable meshes, even though it took me quite some time to grab the idea of the underlying workflow. It does support CSG operations, even though correcting geometrical faults in a somewhat more complex setup is definitely a pain with it.

Regards,

Peter

[hazelsct]

For recent laptops there's also usually some way to get scroll wheel functionality, e.g. my 2007 Dell Latitude uses the right edge of the trackpad for scrolling, and of course Macs use multitouch.