[MtnRunBeachBum]

Good Afternoon Y'all,

The following is partially copied from a post under another forum topic a couple days ago - just trying to get to some additional help, plus I'm consolidating my concerns as I work through the problem. I'm now running a modified, simpler case compared to my original, though at this point the errors appear to be the same...


I'm pretty new as a CFD practitioner, especially OF, so bare with me. (My background is in applied aero, as a test engineer.)

I'm creating a test case where I want to run icoFoam, to keep things simple. I have the model via .stl files, and I'm able to create a mesh via snappyhexMesh.

ISSUE 1: On running checkMesh, I keep getting a high skewness (max=5, 25 highly skewed faces). I've seen some posters mention this is OK. Thoughts?


ISSUE 2: When I try to run the solver, it gets setup to go, loads info from the various dictionaries, then crashes hard. I realize this is the plight of CFD, just looking for some direction on how to proceed. What I think is happening, due to the following in the snappy output, is the mesh cells are not meeting up properly with the model cells:

total attraction master points: 164912 attraction to: feature point: 0 feature edge: 0 nearest surface: 0 rest: 164912
If thats the case (please direct me otherwise), what can I do to "encourage" more appropriate attractions?


Issue 3: Now when I run icoFoam, I get the following:

// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
Create time

Create mesh for time = 0

Reading transportProperties

Reading field p

Reading field U

Reading/calculating face flux field phi


Starting time loop

Time = 0.25

Courant Number mean: 2.41346 max: 2.41346
#0 Foam::error:rintStack(Foam::Ostream&) at ??:?
#1 Foam::sigFpe::sigHandler(int) at ??:?
#2 which: no linux-gate.so.1 in (/home/cfd/OpenFOAM/ThirdParty-2.3.1/platforms/linuxGcc/gperftools-svn/bin:/home/cfd/OpenFOAM/ThirdParty-2.3.1/platforms/linuxGcc/ParaView-4.1.0/bin:/home/cfd/OpenFOAM/cfd-2.3.1/platforms/linuxGccDPOpt/bin:/home/cfd/OpenFOAM/site/2.3.1/platforms/linuxGccDPOpt/bin:/home/cfd/OpenFOAM/OpenFOAM-2.3.1/platforms/linuxGccDPOpt/bin:/home/cfd/OpenFOAM/OpenFOAM-2.3.1/bin:/home/cfd/OpenFOAM/OpenFOAM-2.3.1/wmake:/usr/local/sbin:/usr/local/bin:/usr/bin:/usr/lib/jvm/default/bin:/usr/bin/site_perl:/usr/bin/vendor_perl:/usr/bin/core_perl)
__kernel_sigreturn
#3 Foam::SolverPerformance<double>::checkConvergence( double const&, double const&) at ??:?
#4 Foam::smoothSolver::solve(Foam::Field<double>&, Foam::Field<double> const&, unsigned char) const at ??:?
#5
at ??:?
#6
at ??:?
#7
at ??:?
#8
at ??:?
#9 __libc_start_main in "/usr/lib/libc.so.6"
#10
at ??:?
Floating point exception (core dumped)






So, all ye experts, any suggestions? My main concern I think is with Issue 3, I'm just not experienced, either at OF or as a programmer in general, to be good at deducing where the error messages are trying to send me.

I would appreciate any help or suggestions. Thanks!

[kamakura117]

Hello,

Skewness typically won't cause your simulation to fail so soon, really. The first few places to start are in the definition of your boundary conditions and in the initialization of the simulation.

Your boundary conditions may be non-physical for you simulation, and that will usually cause the problem you're seeing. Sometimes, especially with the pressure boundary in incompressible flow, it just takes a bit of play and some physical intuition. I would start there. You might even consider posting them for us to help you investigate.

Your initial conditions might also result in a pretty spectacular simulation crash. Sometimes it might be more advantageous to relax them and let the simulation build up to the final solution.

Cheers.

[alexeym]

Hi,

Originally Posted by MtnRunBeachBum

Starting time loop

Time = 0.25

Courant Number mean: 2.41346 max: 2.41346
...
#3 Foam::SolverPerformance<double>::checkConvergence( double const&, double const&) at ??:?
#4 Foam::smoothSolver::solve(Foam::Field<double>&, Foam::Field<double> const&, unsigned char) const at ??:?
...

What solver you would like to run?

About your icoFoam log

1. Did you try reducing initial deltaT? 2.5 is a little bit too high for Courant number at the start of simulation.

2. What are the settings of smoothSolver? Do you use preconditioning? What is smoother? Do you set nSweeps?

3. Also, as kamakura117 suggested, your ICs and BCs can be not quite reasonable.

[MtnRunBeachBum]

Thank you all for your input.

In my case, it appeared that the boundaryField definitions were not defined properly. Once I resolved those to properly (I think) reflect a freestream condition, the case ran. I still may not have perfect initial conditions, but at least she runs.

As I mentioned in another post, in my admittedly very brief experience with OF, the lack of any documentation (proper or 3rd party) describing the boundaryField conditions for each type of solver and each variable (U, p, etc.) is a real bummer. I wouldn't be surprised if a lot of new "Foamers" get stuck here without realizing it and turn away. Just a thought.

Overall though, happy with what I'm learning and OF's ease of access to the core of the code. For someone who works in test/applied aero, this really helps to get reacquainted with computational aero.


Cheers!

[reesebl]

Very much this. Even bought the "OpenFOAM Technology Primer," I am pretty disappointed in it as a "Primer" since it really is only cursory in the "how to actually run OpenFOAM" department and spends the bulk of the time on "how to modify OpenFOAM." Anyway, if you have patience and don't mind (a LOT of) trial and error, it is certainly a powerful and useful tool. The documentation though...whew...

I know this is my first post, and don't want to sound negative, I'm running OpenFOAM right now, and I actually kind of enjoy the challenge, but it can the rather frustrating and since this pretty much reflected my feelings wanted to share.

/feelings off

analysis on/

Originally Posted by MtnRunBeachBum

I wouldn't be surprised if a lot of new "Foamers" get stuck here without realizing it and turn away.

[kamakura117]

Hey,

A great place you can start is using an open-source GUI, such as:

http://engys.com/products/helyx-os

This can usually get your simulations started and give you ideas on how to precede with the simulation process. It's a good tool for learning, I think. In reality, many issues you might have are inherently CFD related, not necessarily the code. OpenFOAM just happens to take a bit more effort to find the problems you're having. For example, improper boundary conditions in STAR-CCM+ are equally hard to diagnose, but the GUI makes it pretty easy to iterate, so HELYX-OS might help you in this event. Getting good at CFD is 90% absorbing tribal knowledge. :-)

[MtnRunBeachBum]

lqtm..

Funny you mention Helyx. The larger scope of my project is evaluating a path forward for our company to adopt to bring CFD in-house. Until now its been executed in one-off situations, and the couple people who got smart in it chose OpenFoam due to their programming background. I'm actually leaning towards Helyx as my recommendation at the moment, though still seeing what's out there. It's amazing to me how much more affordable (and usable - even OpenFOAM) CFD has become in the last 5-10 years. Pricing seems to run the gambit from free/low-cost up to $30K+. Pick your poison, as they say.

And to your other points, I do realize my questions are more broad than just applying to OF. My frustrations come from many things I won't get into here.

But to be clear, this really has been fun (when I do manage to figure out my problems) and this forum has been tremendous help.

Cheers!