View Issue Details

IDProjectCategoryView StatusLast Update
0001948OpenFOAMBugpublic2015-12-16 08:35
Reporterriefler Assigned Tohenry  
PriorityimmediateSeveritycrashReproducibilityalways
Status resolvedResolutionfixed 
PlatformGNU/LinuxOSUbuntuOS Version14.04
Summary0001948: icoUncoupledKinematicParcelFoam freezes using cohesionEnergyDensity
DescriptionFor values larger than about 5e9 the code freezes or crashes in case of OF 2.3.x with the following info:
Evolving kinematicCloud

Time = 0.002235

Solving 3-D cloud kinematicCloud
    77 move-collide subCycles
#0 Foam::error::printStack(Foam::Ostream&) at ??:?
#1 Foam::sigFpe::sigHandler(int) at ??:?
#2 ? in "/lib/x86_64-linux-gnu/libc.so.6"
#3 Foam::PairCollision<Foam::CollidingCloud<Foam::KinematicCloud<Foam::Cloud<Foam::CollidingParcel<Foam::KinematicParcel<Foam::particle> > > > > >::wallInteraction() at ??:?
#4 Foam::PairCollision<Foam::CollidingCloud<Foam::KinematicCloud<Foam::Cloud<Foam::CollidingParcel<Foam::KinematicParcel<Foam::particle> > > > > >::collide() at ??:?
#5 ? at ??:?
#6 ? at ??:?
#7 ? at ??:?
#8 ? at ??:?
#9 ? at ??:?
#10 __libc_start_main in "/lib/x86_64-linux-gnu/libc.so.6"
#11 ? at ??:?
Floating point exception
Steps To Reproducerun icoUncoupledKinematicParcelFoam
Additional InformationLIGGHTS works well with that high cohesionEnergyDensity
TagsNo tags attached.

Activities

riefler

2015-12-11 16:11

reporter  

henry

2015-12-11 16:19

manager   ~0005736

Which OpenFOAM version are you running?
You stated product version as 3.0.x but reported the error from 2.3.x.

riefler

2015-12-11 18:16

reporter   ~0005737

I use OF 3.0.x; there is no crash, but the simulation freezes. With OF 2.3.x the same case crashes with the error report above.
Thank you!

henry

2015-12-12 09:11

manager   ~0005738

The issue is with the normalization of the force direction in PairCollision.C

- scalar normalAlignment = normal & pW/mag(pW);
+ scalar normalAlignment = normal & pW/(mag(pW) + ROOTSMALL);

It is not clear what the best form of division stabilization is for this term, it could be done by adding a "small" number to "mag(pW)" or set the result to "0" or "normal" depending on the assumed alignment in the limit of mag(pW) -> 0.

For now I have added ROOTSMALL, if this proves inappropriate please reopen this report.

Resolved in OpenFOAM-3.0.x by commit 27a9f797b9d6d9455f937d6deecf5a0673e7c8f7
Resolved in OpenFOAM-dev by commit 35439753ef974131328ace4fcc1a39aa0bd331eb

riefler

2015-12-13 20:12

reporter   ~0005750

After a git pull and recompilation of lagrangian intermediate and icoUncoupledKinematicParcelFoam, new problems occur. The solver runs through, but the particle disappears and paraview can't open the case (after foamToVtk); setting the cohesionEnergyDensity to zero do not change anything, although usually (OF 2.3.x) the solver keeps the particle and delivers reasonable results.

henry

2015-12-13 20:15

manager   ~0005751

If you make the change

scalar normalAlignment = normal & pW/(mag(pW) + ROOTSMALL);

in OpenFOAM-2.3.x does the case run and post-process?

henry

2015-12-13 20:44

manager   ~0005752

I think ROOTSMALL is too large, I am testing

scalar normalAlignment = normal & pW/(mag(pW) + ROOTVSMALL);

henry

2015-12-13 20:54

manager   ~0005753

ROOTVSMALL is too small, but the case runs with SMALL.

I do not have any problem opening the case after foamToVtk.

riefler

2015-12-16 08:05

reporter   ~0005769

OpenFOAM-2.3.x freezes again with either VSMALL or ROOTVSMALL on both formulas in PairCollision.C with an increased cohesionEnergyDensity (=1e12) at "Time = 0.002235", but OpenFOAM-3.0.x runs now. Still: every solver should sign the particle as inactive at the collision position to indicate that it sticks.
Thanks again

henry

2015-12-16 08:20

manager   ~0005770

> but OpenFOAM-3.0.x runs now

Are you satisfied with the results?

> Still: every solver should sign the particle as inactive at the collision position to indicate that it sticks.

Can you provide a patch for this change?

riefler

2015-12-16 08:25

reporter   ~0005771

> Are you satisfied with the results?
It's all right for me.

> Can you provide a patch for this change?
No, sorry.

Issue History

Date Modified Username Field Change
2015-12-11 16:11 riefler New Issue
2015-12-11 16:11 riefler File Added: icoUncoupled-Fiber+ParticleCollision.zip
2015-12-11 16:19 henry Note Added: 0005736
2015-12-11 18:16 riefler Note Added: 0005737
2015-12-12 09:11 henry Note Added: 0005738
2015-12-12 09:11 henry Status new => resolved
2015-12-12 09:11 henry Resolution open => fixed
2015-12-12 09:11 henry Assigned To => henry
2015-12-13 20:12 riefler Note Added: 0005750
2015-12-13 20:12 riefler Status resolved => feedback
2015-12-13 20:12 riefler Resolution fixed => reopened
2015-12-13 20:15 henry Note Added: 0005751
2015-12-13 20:44 henry Note Added: 0005752
2015-12-13 20:54 henry Note Added: 0005753
2015-12-16 08:05 riefler Note Added: 0005769
2015-12-16 08:05 riefler Status feedback => assigned
2015-12-16 08:20 henry Note Added: 0005770
2015-12-16 08:25 riefler Note Added: 0005771
2015-12-16 08:35 henry Status assigned => resolved
2015-12-16 08:35 henry Resolution reopened => fixed