Main reported bugs (happens only with LIGGGHTS_Flexible_Fibers): using single spheres and multispheres at the same time leads to "Segmentation fault" error / using multisphere integrator delete single spheres.
LIGGGHTS version: LIGGGHTS_Flexible_Fibers 3.8.0
Hello dear LIGGGHTS community,
I am currently facing an issue with the multisphere tool available in LIGGGHTS-PUBLIC.
My main objective is to obtain a mix a single spheres and multispheres within a same simulation and to add breakable bonds between them using the bonds model by Schramm et al. (2003). Up-to-date, this bonds model is available through the LIGGGHTS_Flexible_Fibers version 3.8.0 (see at the end for the git repositery).
To build my model, I took inspiration from different threads talking about dual single spheres/multispheres simulation:
- - https://www.cfdem.com/forums/using-single-and-multisphere-together
- - https://www.cfdem.com/forums/mixing-spheres-and-multispheres
- - https://www.cfdem.com/forums/sphere-and-multisphere-mixing-rotary-drum
However, I meet some issues when I try to use single spheres and multispheres at the same time, especially when I use the LIGGGHTS_Flexible_Fibers version.
To illustrate my words, I built a self-sufficient minimal example, so I invite you to download the three available files attached:
- -
in1.sphere-multisphere-flexible-fibers-issue.txt
: primary example - -
in2.sphere-multisphere-flexible-fibers-issue.txt
: secondary example - -
cluster.multisphere.txt
: multisphere template used in examples
The following development is based on the primary example (in1.sphere-multisphere-flexible-fibers-issue.txt
).
- ~ Using the LIGGGHTS-PUBLIC version 3.8.0, I can easily play with the example without any issue. I can use a particle distribution with only single spheres, only multispheres or a mix of both, only by changing the
fix particledistribution/discrete
command. This works even if both integrators (fix multisphere
andfix nve/sphere
) are activated and associated to each multispheres or spheres groups, whatever the order of integrators activation (fix multisphere
beforefix nve/sphere
or vice-versa). - ~ Using the LIGGGHTS-PUBLIC-JKR version 3.8.0, the result is exactly the same and there is no issue to report.
- ~ Using the LIGGGHTS_Flexible_Fibers version 3.8.0, first issues appear, depending on the particles defined in the
fix particledistribution/discrete
command. - → Only single particles → OK but only if the multisphere integrator is deactivated. If the multisphere integrator is activated, the generated single particles disappear.
- → Only multispheres → OK, no issue to report, even with both integrators are activated.
- → Both single particles and multispheres → ERROR. After the insertion of particles, computation crashes and leads to a "Segmentation fault" error message:
[aar096:32167] *** Process received signal ***
[aar096:32167] Signal: Segmentation fault (11)
[aar096:32167] Signal code: Address not mapped (1)
[aar096:32167] Failing at address: 0x20287a4e0
[aar096:32167] [ 0] /lib64/libpthread.so.0(+0xf630)[0x2b11d2b5c630
I also explored several leads:
- - Running on only one core.
- - Replacing single particles by a multisphere of one particle to get around the problem (but the particletemplate/multisphere fix command prevent the use of nspheres equal to 1).
- - Separating both particles generations (as proposed in
in2.sphere-multisphere-flexible-fibers-issue.txt
: single spheres insertion and then multispheres insertion), but the presence of multisphere integrator delete the single particles as soon as they are inserted. - - LIGGGHTS-PUBLIC → https://github.com/CFDEMproject/LIGGGHTS-PUBLIC.git
- - LIGGGHTS-PUBLIC-JKR → https://github.com/eidevag/LIGGGHTS-PUBLIC-JKR.git
- - LIGGGHTS_Flexible_Fibers → https://github.com/schrummy14/LIGGGHTS_Flexible_Fibers.git
Consequently, it seems not possible to make a simulation with both single spheres and multispheres using the current LIGGGHTS_Flexible_Fibers version 3.8.0. It seems to come from the fact that, in this LIGGGHTS version, the activation of multisphere integrator (fix multisphere
) deletes single spheres. This is my main hypothesis, but I am not 100% sure.
For now, I am running out of ideas, so I leave it to any other good intuition of experts.
Links to LIGGGHTS version studied:
Thank you very much in advance ! ☺
Best regards,
Theo
Attachment | Size |
---|---|
![]() | 2.77 KB |
![]() | 3.02 KB |
![]() | 94 bytes |