hex to wedge elements

[tstanley]

On occasion a situation arises where it would be nice to join a single face of a hex element to two wedge elements so that all the nodes are merged at the interface, but one edge of the wedge elements crosses the face of the hex element. What sort of inaccuracies in the results would occur from doing this.

Tom

 

[aerostress82]

If you are using 1st order elements, and it is a simple linear static analysis, probably no significant error for other locations of the FEA.
But if you want to look at the stress distribution around these regions, you will see weird outputs because of that weird edge.


If you are using 2nd order elements, your analysis results will be wrong as you will now have an additional node in the middle of that edge, and that node will now be free to deform without being supported by the rest of your FEA. In addition to this, what I mentioned above about weird stress distribution (plots) is also valid for this as well.

Spaceship!!
Aerospace Engineer, M.Sc. / Aircraft Stress Engineer

 

[tstanley]

Thanks for your reply.

I am only using first order elements and all the nodes are connected. The model that I am working on is a unit cell for a twill weave composite and the problem only occurs in the matrix, not the fibers. If I refine the mesh I will still end up with the same problem. I could probably use a tet mesh but would like to avoid that. In the end my goal is to come up with a reasonable estimate of the young's modulus and shear modulus of the ply. I have been speculating to myself that the error would end up being like a defect in the matrix. Could there end up being a gap between the elements under load where the faces meet?

I have done this type of model before simulating a plain weave composite and got reasonable results when compared to a physical test where the plain weave was one ply in the layup. I did not really see any weird outputs but they might have been hard to see because of their location in the model.

It is the unknown that worries me.

Tom

 

[aerostress82]

I assume this is happening at the lovation where the matrix is touching the laminate ply.
If this is the case, I understand your unknown and raise ypu one.

In theory, the matrix will support out of plane forces, and there will be some interlaminar shear at the matrix faces. And then you will have in plane loads mainly supported by your ply. Everything looks good ao far. But when the FEA matrix is created, the matrix array will slightly be dofferent than that non-congruent mesh area. Same number of nodes, yes, but the physical effect of this slight difference between congruent and non-congruent element edges is still questionable.

I would advise you to do a workbench with this other bersion too (with a tet as you named it) and then compare the stresses and strains at those ply elements where mesh gets weird. You should have the answer. A mathematician could help too, but he/she should be aware of the stiffness matrix compilation for conposites-which will be hard to find.

So only solution is a conparison of atresses and strains by a workbench, or someone else here who already performed this workbench and know the results already. Good luck. Please let me know if you need anything else on this.

Spaceship!!
Aerospace Engineer, M.Sc. / Aircraft Stress Engineer

 

[tstanley]

Thanks again.

In the end I will take your advice and do a tet mesh, but it will be a while before I get there.

Tom