construct parallel bond model in yade

Asked by steve on 2020-10-05

hi everyone,
I am trying to construct a parallel bond-model to simulate concrete in yade. it is a model that uses 'short beam' connecting two particles so that it can transfer both force and torque. But after reading several similar questions (eg. #635871 and ## 661250), it seems that none of the JCFpmMat or the CPMmat is suitable for the parallel bond. From my point of view, the only difference between the JCFpmMat interaction and the parallel bond model is that the parallel bond model can transfer torque as well.
so my question is:
1. is it possible to construct the parallel bond model by allowing moment transfer in JCFpmPhys?
2. is it possible to add two interactions ( both jcfpmmat & cohfrictmat)?
2. in the source code of JCFpmPhys[1], line 655 and line 656, I can see that 'crossSection(pi*R^2)' take place the' R1*R2/(R1+R2)', why is that? if I was given the modulus of bond (E), how do I compute the jointNormalStiffness(jkn)? is it jkn=E or jkn=E/R?
thanks and regards

Question information

Language:
English Edit question
Status:
Solved
For:
Yade Edit question
Assignee:
No assignee Edit question
Solved by:
Robert Caulk
Solved:
2020-10-11
Last query:
2020-10-11
Last reply:
2020-10-06
Best Robert Caulk (rcaulk) said : #2

1. is it possible to construct the parallel bond model by allowing moment transfer in JCFpmPhys?

Yes, it is open source code you can modify/add/subtract anything you want. Relevant thread [1].

2. is it possible to add two interactions ( both jcfpmmat & cohfrictmat)?

Do you mean add two laws? Each of these follow a different law, the law decides how the forces are determined, when the bonds break, how the plasticity develops. No, you cannot use two laws on a single interaction.

2. in the source code of JCFpmPhys[1], line 655 and line 656, I can see that 'crossSection(pi*R^2)' take place the' R1*R2/(R1+R2)', why is that? if I was given the modulus of bond (E), how do I compute the jointNormalStiffness(jkn)? is it jkn=E or jkn=E/R?

Please, in the future, add the links, don't make us wade through lines of code just to end up *maybe* at the same place you are referencing.

It would appear that they changed from the harmonic mean to the cross sectional area to keep the expression size independent. I realize that doesn't provide much extra information, but from a units stand point - it is true. Size independent expressions are always preferable in DEM. The user defined Normal stiffness [2] should be in units of Pa/m to stay consistent with this expression. So I suppose you would use jkn=E/R, to answer that question.

Cheers,

Robert

[1]https://answers.launchpad.net/yade/+question/689058
[2] https://yade-dem.org/doc/yade.wrapper.html#yade.wrapper.JCFpmMat.jointNormalStiffness

Jérôme Duriez (jduriez) said : #3

In case this helps, I would emphasize that jointNormalStiffness macroscopically relates to an interface stiffness (hence the Pa/m), not to a bulk stiffness (in Pa), like the Young's modulus.

steve (steve-cash) said : #4

Thanks Robert and Jerome for your kind reply! and apologize for not attaching the links. question 2&3 are clear now.
for the question1, Yes, I have looked at the Cohfrictmat. but it seems that the CoheFrictPhys only defines one stiffness for both direct-contact and joint while in the parallel bond model I desired there should be separate values for the joint and contact. After I read the thread you referred(#689058), it seems that the moment and twist doesn't have a obvious effect on the macroscopical results.
however, I do want to make modification on the source code to take account the moment and see its effect on the outcome. But due to my limited programming ability, I dont know how to achieve that, the idea is: obtain the relative rotation angle-->calculate the moment by M=Kn*I*angle, where the I is the moment of inertia which is some constant times pow(R,4) in this case--> calculate the stress based on classic beam theory( normal stress due to bending = M*R/I)-->check if the total stress(F/A+M*R/I) exceed the tensileStrength. Anyway, I think I'll start with JCFpmat for now.
in the end , may I conclude that there is no exact parallel bond model in yade currently which requires different stiffness values for direct contact and cohesive bond while allowing the moment transition. but maybe JCFpmmat can achieve similar results without moment transition or you can modify the source code to get the exact model?

Robert Caulk (rcaulk) said : #5

>in the end , may I conclude that there is no exact parallel bond model in yade currently which requires different stiffness values for direct contact and cohesive bond while allowing the moment transition

I hope it is clear that a joint != a broken bond in JCFpm. Worth reading [1] if you were mistaken there.

[1]https://www.sciencedirect.com/science/article/pii/S1365160912000391

steve (steve-cash) said : #6

Thanks Robert, I 'll have a look at the article. I'll post here in case I have further questions regarding this.
cheers,
steve

steve (steve-cash) said : #7

Thanks Robert Caulk, that solved my question.