How to connecting solid interface meshes together?

[bjpil]

Hello all,

I am having troubles connecting two solid element meshes together in MSC/Nastran. For simplicity sake, I have two cubes, one on top of the other with a small gap between them, meshed using hex elements. At the gap interface, I connected rbe2, constraining all translations. The bottom nodes of bottom cube were constrained in translation to prevent vertical (Y) and normal (into the sheet) movement (X). The left sides of both cubes are constrained to prevent horizontal movement. A force was then applied to some of nodes in the normal direction (X). I was hoping to find the shear stresses at the interface (gap) with this method but I end up with a rotation (MAX PIVOT) problem at the interface. What is a better method, excluding equivalencing the interface nodes, to prevent the rotational problems?

Thanks,
BJP

 

[pjhype]

The reason for the singularity is not obvious based on what you described. I am assuming that the mesh on the upper cube matches the mesh on the lower cube one-to-one across the interface. I am also assuming that you are using AUTOSPC. You might try constraining the independent nodes on the RBE2 element in the rotational degrees of freedom and constraining all degrees of freedom on the dependent nodes to the independen nodes by the RBE2's.

When you make a connection with RBE2 elements in the manner that you have, you are essentially equivalencing the interface nodes. Is there some reason why you do not want to equivalence the nodes? If the cubes are bonded together with an adhesive, there are other considerations (such as boundary layer effects) that will cause the interface shear stress to (mathematically) approach infinity at the interface. The use of rigid elements at the interface is not adequate to model this phenomenon, however.

pj

 

[bjpil]

Hi PJ,

What I was trying to do was set up a cantact analysis. I was trying to control the horizontal movement with the RBE2s. I was going to assume that there was no slipping at the interface. Becasue I was using a non-linear solver there was no AUTOSPC. What I ended up doing was constraining all rotational degrees of freedom for all the solid elements. This solved the problem but I did not want to initially do this because I didn't, and still don't, know what kind effect it would have on the solution.

I then scrapped this method and decided to solve the problem using a couple of soft springs, turning on friction and solving first for the vertical load, that activated the friction, and then applying the vertical load. I still needed to constrain the rotations.

I also thought that connecting the RBE2s in the manner that I did was to simulate equivalenced nodes but it does not. I guess the rotation degrees must be constrained.

Thanks for your suggestions,

BJP