Stress averaged between different materials

Sergio

Hi Victor,

Very often I model different materials that are glued, but for simplicity I model as a continuous mesh and make two different groups of elements for each material, sharing the nodes at the interface. Problem arise when I postprocess as the stress looks like if it were averaged between nodes of different material at the interface. I'm  not sure if this averaging is valid or not, but take a look at this example and how he has separated the materials by means of equations, having then the stresses separated for each material:

https://github.com/mkraska/CalculiX-Examples/tree/master/NonLinear/Sandwichtest

Is possible to add this in future versions of Mecway? In others postprocessors I have an option to average or not the stress of different materias. Take a look in this example, in the right side of the results, how the stress in the first column of elements of the grey material is mixed with the external material.

Regards

Victor

I see. It's too minor for including any time soon though. There's also the converse problem where it should average at boundaries like bonded contact with the same material and symmetry planes.

mkraska

Stress averaging across material boundaries is not correct. Stresses can have discontinuities at such boundaries. I don't see the point of averaging at symmetry planes. What is there to average? Both sides would provide the same value. I wrote a script which completely suppresses averaging by providing separate nodes for all elements touching an existing node and adding appropriate equations. https://github.com/mkraska/CalculiX-Examples/tree/master/Linear/Separate Martin

Victor

I agree that not averaging them would be more accurate. With the constraint of node values being points on a single-valued function, averaging seems like the best it can do. In future, I might have to add a new kind of mixed field variable type that has averaging at some nodes and multiple values at others. But this would be quite a big job so again low priority. There are cases where stress should not be averaged at material boundaries too so that probably has to be an option for each boundary or material pair.

Thanks for posting the no-averaging script for CCX. Note that the internal solver does this already. The problem there is that it should really show both averaged and non-averaged values at their appropriate places in the same contour plot.

I also agree with you about symmetry for stress. I was probably thinking of heat flux. Somebody was once confused why there was a small non-zero heat flux normal to a symmetry plane or insulated surface. If that was averaged with the opposite side having the opposite sign, it would become zero.

Sergio

Welcome to the forum, Martin, your examples are very usefull. Normally for this cases (rubber parts) I don't include the metal part as they are not the matter of the analysis, sometimes I need it for contact or other purposses.

But as you said, nodal averaging is not correct and maybe with similar stiffness materials the problem is not so apreciable, but for soft ones mixed with rigid does it. I use TIE to join the surfaces, but sometimes the TIE doesn't work well and get some nodes going over the rigid surface, that's why I want to investigate the equations.

I repeat the example separating the materials (using Mecway's Disconnect element feature and joining with TIE), in the right side, and we can see the difference in stress in the same node, almost twice the computed stress for the averaged (left side).

Sergio

Guess that the Disconect Element feature and TIE is good enough procedure for getting accurate result at the interface nodes From my point of view there is no need to worry in this matter.