Cyclic symmetry component

[Anonymous]

A few days ago I had to verify some FEA results that a manufacturer send us of a rotatory component (they use ANSYS). This particular component had a cyclic symmetry that allows to perform the analysis of a 1/10th portion of the entire thing. This component is going to be subjected to both angular velocity as well as compressive loads, so it is necessary to check both static linear stress as well as buckling (just critical buckling load - linear-) and it can be modelled with a combination of shell and solid elements (discs, radial arms and a core). So, I began to set my boundary conditions and I encounter my first problem. Autodesk simulation mechanical (ASM) does not allow to apply a cyclic symmetry condition to an edge (because of the cyclic symmetry, the discs - shell elements- get "sliced" so a symmetry surface would have to be the discs thickness) so I came up with a solution that I'm not sure if it is a correct one. I added to the CAD model some surfaces perpendicular to those cut edges with a "height" of the thickness of the discs (see attachment 1). I thought that in this way I was not affecting the stifness (and the overall configuration of the component for that matter) and that it would allow me to apply the cyclic symetry to those surfaces. I did not get the results I was expecting specially when I added the angular velocity boundary condition (Omega). So I began to wonder if I was properly setting the cyclic symmetry condition and decided to do an experiment that I would like to share with you hoping that someone can help me clarify this issue.

I crated a simple geometry with cyclic symmetry and that can be modelled with a combination of solid and shell elements (just like the component I had to check). The I ran three analysis: one for the complete component, one for half a model and another for a quarter of the model as follows: First a static stress analisys considering only angular velocity. Then a static stress applying only compressive forces. And finally a critical buckling load with that same compressive force using a frictionless constraint in the simmetry faces because this type of analysis doesn't allow neither cyclic symmetry nor local coordinate systems. For the portion models I used cyclic symmetry (recreating the "additional thickness surfaces" trick I mentioned earlier) and then I tried with regular symetry condition in the tangential direction with a local cylindrical coordinate system (I think that this last aproach is not appropriate becouse that restriction wont allow any rotation of the discs relative to the core and that is not realistic in this particular case, but I did it any way just for comparisson) Some of the graphical displacement results are presented in the attachment 2. This experiment results showed me that in fact there is something I'm doing wrong or something I'm missing in the set up of the model because I obviously wasn't able to reproduce the complete model results neither for the half model nor the quarter model specially for the omega ones. Here are some conclusions from the experiment (you can see the images in the attachment number 2).

The results for the displacements when applied only angular velocity where completely different for the three models.

The displacement results for the only compressive loads models where relatively similar between the three models. It was kind of surprising that the half model results were so different from the cyclic symmetry one being that it is in fact a "local Y" simmetic model

The critical buckling load multiplier is kind of similar for the three models but because of the inability to apply cyclic symmetry boundary condition in a buckling load linear analysis, and thus preventing the relative rotation between the discs and arms relative to the core I don't know how much of "coincidence" factor there is for this results. Althow it appears to be a very similar displacement pattern

So, If there is anyone still reading at this point (I know is a long post), is there some fundamental concept I'm not familiar with? has anyone stumble upon something similar?

By the way: I created all the models meshes with an absolute mesh size value and deactivated the mesh relative to size parts in order to obtain similar meshing for the three models. I also check that I applied identical boundary conditions both for constraint and loads

Thank you all very much. 

[Anonymous]

Hello again

So I went ahead and keep on trying to figure out what is going on. So I went down to the available tutorials and documentation. I started to wonder if there was something "peculiar" about the geometry I'm trying to analyze. There are several differences that I can point out. The common geometry that is shown using this type of restriction (cyclic symmetry) is often a fan (blades plus central core) with the symmetry planes or surfaces splitting the core. That is the first big difference that I think is giving me trouble because the blades are "free" to move, or at least their displacement is not greatly affected by the cyclic symmetry boundary condition (This is also why I think the "frictionless" restriction applied on those surfaces is an appropriate assumption because of the very small tangential displacement of the core relative to the blades.) The component I’m analyzing has a different behavior in the sense that the arms of the structure and the discs have in fact a tangential stiffness depending on the cyclic symmetry condition. The other difference is the fact that the cyclic symmetry is always applied on surfaces of parts modeled with 3-D solid elements. So I tried a new approach and added solids on the edges I needed to apply the cyclic symmetry condition (see attachment). That exercise gave me great results. As you can see from the attachment 2, the displacements where the same between the complete and the quarter model. I thought I had the answer , but then I tried to analyze the complex geometry with the same approach and it gave me some poor results. The displacements where not even near what I was expecting. So… the only approach that was left (or at least the only one I can think of)…. I had to model the entire thing with solid elements. I used a layered mesh instead of the plate elements and I was finally able to get the results I was expecting but, of course, using with far more elements compared to the solid-shell model. I still do not understand why my previous approach (previous post) did not work and I do not understand why ASM does not allow cyclic symmetry boundary condition only to surfaces. Any way… I thought I should post this in case someone in the future can use this information or in case someone comes up with different approach... or even maybe someone can tell me what I’m missing.

 

Thanks a lot…