Open Channel analysis will not complete

[ForrestJudd]

I'm trying to simulate water flowing from a canal into a pipe.  I've tried many different loading situations, time steps, mesh sizes, etc. and I just can't get the simulation to complete.  In almost every case I get a message that the analysis stopped with 1 warning and 0 errors.  The warning is almost always that the time step is too small.  I understand why that might be, but I've had other analyses complete despite that warning.  I don't see anything in the analysis results that would indicate the problem.

The archive is about 2.5 MB and is available here:

https://skydrive.live.com/redir.aspx?cid=48cb900016e0723f&resid=48CB900016E0723F!250&parid=48CB90001...

If you have issues with SkyDrive, you can also download it here.

I have two different scenarios in this one, with gravity pointing two different directions.  Neither of them will complete.  Any help with this would be greatly appreciated.

[ForrestJudd]

Ok, I think I'm missing something fundamental here.  I did the manometer exercise from the official 2012 courseware.  It looks like the attached image.  That analysis ran successfully.

I try to do the exact same thing with a nearly identical model, and I get the problem I mentioned above - the analysis fails.  I'm convinced now it has something to do with my mesh - the mesh in the courseware exercise is very clean, and looks to have been manually created.  I'm using the mesher (and I've tried various settings all to no avail).  I end up with elements in the bends of the manometer like the second attached image.

Is the solver really that sensitive to the mesh?  If so, how can I reliably get a mesh that won't lead to failure when I'm using complex geometry?

[ForrestJudd]

Another update - I created a mesh manually of just a long cylinder with a small initial volume, and I was finally able to get an analysis to complete. I created the mesh by dividing a 2" circle using 8 divisions, then created a MRS mesh 50" long with 100 copies (leading to a .5" vertical element dimension).  I ran the analysis over 1.5 seconds using 200 steps.

I created a cylinder with identical geometry in Inventor, with the same initial volume, analysis parameters, and boundary conditions.  This analysis failed after 21 steps.

I really hope I'm missing something fundmental here, otherwise the open channel analysis type doesn't appear to be at all useful.

[John_Holtz]

Hi Forrest,

It sounds to me as if the mesh is too coarse with your CAD models. In your hand-built model (and the manometer example), you are quaranteed to have the same number of elements in the cross-section along the entire length. With a CAD model and an automatic mesh, you are somewhat "at the mercy" of the solid mesher; that is, you do not have control over the number of elements (really the number of nodes) in the cross section. If there are too few nodes at a particular cross section, the flow can become chocked and difficult to converge.

Did you use the boundary layer mesh with the CAD models? That will put N number of elements at the wall and can help ensure there are at least a given number of nodes in the cross section. Please see "Getting Started > Help > In-Product Help" "Help > Autodesk Simulation > Mesh Models > Mesh Overview > Meshing CAD Solid Models > Model Mesh Settings > Solid".

[ForrestJudd]

John,

Thanks for the reply - I really appreciate it.  I don't think this is a problem with the number of nodes.  I took the simple cylinder example and decreased the mesh size significantly and I end up with the same problem.  I don't think the analysis log is indicating lack of convergence - when it stops at a given time step, it hasn't reached the maximum number of iterations specified in the analysis parameters.

I've attached a screen capture of results when the analysis stopped, with nodes selected to give you an idea of their density.  I've also attached the analysis log.