CFD run keeps failing


#1

I’m running a vehicle CFD simulation. I used the FSAE webinar as a guideline for setting the BC and the numerics. I used the hex-dominant parametric mesh and the simulation keeps failing. When using the wind tunnel mesh the simulation ran fine and converged. This is what the error log says at the end:[14] Writing current time 2
[15] Signal 15 encountered
[15] Resetting old handlers (just in case)
[15] Unset SIGFPE(8) signal handler
[15] Unset SIGSEGV(11) signal handler
[15] Unset SIGTERM(15) signal handler
[15] Unset SIGQUIT(3) signal handler
[15] Writing old times:
[15] 1 times to write
[15] Write t=1 to “processor15/1”
[15] Writing current time 2
[8] Signal 15 encountered
[8] Resetting old handlers (just in case)
[8] Unset SIGFPE(8) signal handler
[8] Unset SIGSEGV(11) signal handler
[8] Unset SIGTERM(15) signal handler
[8] Unset SIGQUIT(3) signal handler
[8] Writing old times:
[8] 1 times to write
[8] Write t=1 to “processor8/1”
[8] Writing current time 2
[9] #0 Foam::error::printStack(Foam::Ostream&) at ??:?
[9] #1 Foam::writeOldTimesOnSignalFunctionObject::sigHandler(int) at ??:?
[9] #2 in "
[9] #3 Foam::GAMGSolver::scale(Foam::Field&, Foam::Field&, Foam::lduMatrix const&, Foam::FieldField<Foam::Field, double> const&, Foam::UPtrList<Foam::lduInterfaceField const> const&, Foam::Field const&, unsigned char) const at ??:?
[9] #4 Foam::GAMGSolver::Vcycle(Foam::PtrListFoam::lduMatrix::smoother const&, Foam::Field&, Foam::Field const&, Foam::Field&, Foam::Field&, Foam::Field&, Foam::Field&, Foam::Field&, Foam::PtrList<Foam::Field >&, Foam::PtrList<Foam::Field >&, unsigned char) const at ??:?
[9] #5 Foam::GAMGSolver::solve(Foam::Field&, Foam::Field const&, unsigned char) const at ??:?
[9] #6 Foam::fvMatrix::solveSegregated(Foam::dictionary const&) at ??:?
[9] #7 Foam::fvMatrix::solve(Foam::dictionary const&) at ??:?
[9] #8
[9] at ??:?
[9] #9
[9] at ??:?
[9] #10 __libc_start_main in "
[9] #11
[9] at ??:?


mpirun noticed that process rank 9 with PID 162 on node exited on signal 15 (Terminated).

I have no idea what any of this means so if anyone could help I’d appreciate it. The project is private so if you want to see it let me know


#2

Hi @vpatel00 - could you share the link to the project / make it public for us to take a look?


#4

Sorry, I didn’t see the make ‘public’ option. I’ve made the project public now, here’s the link:
https://www.simscale.com/projects/vpatel00/sem_car_c3_1/


#5

Hi @vpatel00,

Do keep using mesh 2 as it has zero illegal cells and will give you less problems. You have done a simulation with mesh 2 in run 3 and it has progressed the furthest, so keep using that setup.

The instability can be due to either the low mesh quality (even with zero illegal cells) or the numerics.

I suggest fine-tuning the mesh using the hex-parametric option. You have done so once for mesh 1 but that has produced a lot of illegal cells rendering the mesh unusable. This could be caused by the geometry where you seem to have a outer and inner shell intersecting each other. You want to remove the outer shell and hence remove the intersection so that the mesher can cooperate better. Ensure good CAD practices are applied and your geometry has to be as clean and defect free as possible.

The other solution which I suggest you try first is to adjust the numerics in terms of the gradient scheme for default, grad§ and grad(u) to cell limited Gauss linear and see if manages to be stable. If this does not work then you will need to perform the first solution towards the geometry as mentioned earlier and re-mesh.

Also note that mesh 3 should preferably not be used unless it is improved on as it has over 150 illegal cells. Note that the layering for the bounding box is in the wrong direction and should be at the floor rather than the side in order to better capture ground effects. You also will need to assign a boundary condition (no-slip wall) to your car geometry and the result control as well to that same geometry so you can monitor it for convergence and obtain some preliminary data.

Cheers.

Regards,
Barry


#6

Hi Barry,

Thanks for the reply. Mesh 2 was the automated wind tunnel mesh option which is not fine enough for my needs. Thanks for spotting the incorrect bounding box. I didn’t assign the no-slip BC on the body as I was lazy but will do so. I thickened the surface in the cad program in order to import the geometry in Star ccm. I’ll try to import a new geometry without thickness and see if this solves the mesh issues.