Problem to simulate a simple compression test

JF's picture
Submitted by JF on Thu, 07/08/2010 - 15:30

Hi everyone,

I try to make a simple compression test (see different attached files) from a cohesion sphere packing. Therefore, the top wall modeled by mesh is moved down.

But I have a strange simulation (you can run it from the attached file), there are not stresses computed (see ./post/Stress_file-*) on the top wall. Moreover the top wall only push some spheres.

I am a beginner, so it is possble that I done a mistake. if anyone can explain me it, I will be fine for me.

Thanks
Jf

Ps: I know that the chosen time step is too high but it cannot explain these problems.

AttachmentSize
Binary Data compression.tar_.gz9.56 KB

Forums:

ckloss's picture

ckloss | Thu, 07/08/2010 - 17:18

Hi JF,

a couple of comments:

1) in order for the stress to be really calculated, you need to use fix mesh/gran/stressanalysis, not mesh/gran (mesh gran - I will issue a patch (1.1.5) to add a warning to the code that is issued if the wrong mesh/gran is used

2) moreover, in your simulation there is only one series of particle-wall collisions, the duration is just a few timesteps (around timestep 12500) - and only during particle-wall collision, stress is generated. By dumping every 1000 time-steps, you don't see stress in the simulation for this reason

3) The thing with the spheres not influenced by the mesh: This is actually a numerical issue caused by the fact that the sphere positions are exactly at the edge of the triangles - due to roundoff/machine precision no contact is calculated for both tris - there is a numerical parameter to account for this. So far, I thought it would be appropriate as is - Now I will adapt it

4) one more note: I am sure you are aware that the particles are _extremely_ soft - I dont think the contact model is still valid in this parameter range

So try version 1.1.5 (to be released soon) and let me know if it does what you expect

Cheers,
Christoph

JF's picture

JF | Fri, 07/09/2010 - 07:57

Hi Christoph,

Thanks for your very quick reply. Concerning your comments:

1) Your are right, it's my fault. I done a lapse in concentration, I forgot the line: fix mesh/gran/stressanalysis

2) I understood your comment

3) ok for the numerical parameter.

4) Yes I used a low Young moduls in order to use a high value of time step. But soon, I will have to add a contact law (e.g., pair_gran_elastic.cpp) within LIGGGHTS and to do it I'll bother you. Sorry ;-) .

Regards
Jf