Someone in the forum of calculix dropped this as reference.
From this I am assuming the ux is translational displacement field and Urx rotational field.
Are they cancelled in distributed coupling?
I am a begginer so i am trying to understand what is behind.
Also is it correct to use distributed coupling? How can I know it is better fit to my solution?
This thread is indeed a good place to start. It explains the difference between kinematic couplings (introduce rigidity like rigid body constraints) and distributing couplings. There are also several blog posts showing this difference in Abaqus. For example:
The table only shows what can and what can’t be applied to the reference node. So with *COUPLING followed by *DISTRIBUTING, you can only apply forces and moments, not displacements and rotations.
Distributing coupling is the right choice if you don’t want to make the (sub)surface rigid.
As you can see from the attached examples (and in your model) the selected subsurface can bend in one model and cannot in the other model which may indeed affect the displacements significantly. Kinda like this (MPC PLANE):
I commend you for your investigation to understand the differences in kinematic and distributed couplings. If you will allow me, I would like to add a few general observations about these.
First, while there are times when you do care about getting the better answer in this region, this is not always the case. Incorporating distributed couplings everywhere because it may give a more realistic answers in many places is not always necessary. Saint-Venant’s principle tells us that, if the results we seek are for determining margins far away from this interaction, then how we model this interaction is not so important. And we would typically like to keep the model as simple as reasonable and would like to minimize solve time.
Second, I point out that distributed couplings likely give us a better answer locally, but it is still not giving us an accurate prediction of reality. The results are simply less conservative than when using the kinematic coupling.
Finally, I recommend that you consider whether you can run your models with the kinematic couplings first, just as you have done here. As you have shown, the stress results are intuitive, and they build confidence before changing to distributed couplings in only the regions of high stress. The results in these regions with distributed couplings will not be as intuitive.
Thanks a lot. I’ll try to keep in my mind your advices. But for me it is harder to understand which results in which case represents close results in reality.
Do you recommend books or references about how we approach the results in fem analysis? Also the techincs and hints for it. I know the basis of the fem in terms of formulations but for example this kind of coupling staff is next level for me. @FEAnalyst Also your comments are valuable for me in this case.
Do you mean edge of a solid model or just 1D (beam) elements ? This constraint is applied to a surface and this surface must be based on element faces. Of course, 1D and 2D elements in CalculiX are internally expanded to solids, but their face numbering is different and they don’t work well with some constraints (e.g. shells with rigid body constraint).
I would mainly recommend the CalculiX User’s Manual and the CalculiX forum for such topics. Apart from that, Abaqus documentation is sometimes helpful (if you keep in mind the differences between these solvers). There are some good books too, but mostly about other software or just FEA in general. I can definitely recommend Dominique Madier’s book. He uses Nastran and describes such constraints in one of the chapters. Kinematic coupling corresponds to Nastran’s RBE2 while distributing coupling is equivalent to RBE3.
I mean on solid part there is a surface to apply forces, I want to apply the forces on the edges. I want to apply distribtuing coupling on a surface edge on the solid element.
It’s better to avoid applying forces and BCs to edges in 3D solid models because this may cause artificial stress concentrations. Instead, it’s recommended to use a small surface patch.
Anyway, surface traction load can only be applied to faces and the distributing coupling constraint that you use needs faces too. However, there are other options such as concentrated force load (just divide the magnitude by the number of nodes) and rigid body constraint or other couplings that can be applied to node sets instead of surfaces.
Sometimes, it even makes sense to model at least a small portion of the structure that transfers the loads to the actual component of interest via contact or bonded interface. Or to use spring/GAP elements as kind of flexible support.
