Yade

Twisting moment

Asked by Luc Sibille

Hi,

I would like to use the Law2_ScGeom6D_CohFrictPhys_CohesionMoment contact law with only bending moment and whitout twisting moment. An easy way would be to assign alphaKtw = 0 in order to obtained the twisting stiffness ktw = 0. However due to harmonic average :
2.0*sdec1->alphaKtw*sdec2->alphaKtw/(sdec1->alphaKtw+sdec2->alphaKtw)
it is not possible to assign alphaKtw = 0 for both particles involved in the contact. Consequently I would suggest to add in Ip2_CohFrictMat_CohFrictMat_CohFrictPhys a test as:
if sdec1->alphaKtw==0 and sdec2->alphaKtw==0 then ktw = 0

Would you agree with this suggestion? would you suggest something else?

By the way, the twisting moment leads to crazy spin velocities, but I have not investigated whether it is "normal" or whether it is related to a bug.

Best,
Luc

Question information

Language:
English Edit question
Status:
Solved
For:
Yade Edit question
Assignee:
No assignee Edit question
Solved by:
Luc Sibille
Solved:
Last query:
Last reply:
Revision history for this message
Bruno Chareyre (bruno-chareyre) said : 		#1

> if sdec1->alphaKtw==0 and sdec2->alphaKtw==0 then ktw = 0
It makes sense.

> By the way, the twisting moment leads to crazy spin velocities, but I have not investigated whether it is "normal" or whether it is related to a bug.

Twisting/bending are not included in global particle stiffness in the
timestep determination, if they have dominant effect the automatic
timestep may be too large.
If you show that the same problems occures even with smaller steps then
it could be a bug. Would be surprising though, since twist moment have
been used many times without problems.

Bruno

Revision history for this message
Luc Sibille (luc-sibille) said : 		#2

You're right Bruno, crazy spin velocities are related to instability of the explicit time integration scheme in rotation, a smaller time step solved the problem. This may be misleading since in this case the translational motion stays stable and the particle assembly does not "explode".

Luc

Revision history for this message
Bruno Chareyre (bruno-chareyre) said : 		#3

It is not difficult to include these moments in the timestep determination from a mathematical POV.
The problem is programming it: not all interactions have rotational terms, thus we would need dynamic casts.
Painful to implement efficiently, but if someone is willing to do that it would be a nice improvement.