Learning Forum

[Jana Rittelmann]

I cut my model so that it represents an eighth of my actual geometry and used cyclic symmetry to get the results on the full model. Unfortunately, I get high stresses along the edge of the face I used as low/high boundary. 

It is important to note that this edge is part of a contact region and that I do expect high stresses in that region. I tried modelling a quarter model and the high stresses went away for the region that I didn't cut (i.e. every other edge from the original eighth). Ideally, I would just model the full geometry, but unfortunately that would take very long to solve and I would like to avoid that. 

Is it not valid to define cyclic symmetry in contact regions/regions of higher stresses then?

And can I safely assume that the stresses in my quarter model are correct (in the region where I didn't cut compared to the eighth model)?

[Ashish Khemka]

Hi,

The stress in the quarter model should be correct but did you check the stress level at the same interface in quarter model and eighth portion of the model? Also, if you can share the snapshots then it might be easier for forum members to comment.

Regards,

Ashish Khemka

[Jana Rittelmann]

 

Hi Ashish, 

thank you for your answer. Maybe the following snapshots will provide more clarity: 

This is the full geometry, eighth and quarter geometry in cross section, respectively: 

   

The inner part rotates inside of the outer part. Now, I want to look at the stresses on the corners of the octagon (inner part). Due to contact with the outer part, I do expect higher stresses there. 

But as you can see in the following pictures, there is a line of high stress along the edge of the cyclic boundary: 

If I take the quarter geometry and look at the corner that has not been cut, it looks like this (it is the same location as before): 

I also looked at nodal differences for the two versions and they are much lower for the corners in the quarter model that have not been cut. 

From this, I would assume that for some reason (maybe because of load/contact?) I cannot cut the corners like I did with the eighth model. But as suggested in your answer, Ashish, I hope that for the quarter model, the stresses should at least be correct for the uncut corners?

Thanks in advance! 

 

[Ashish Khemka]

Thanks for sharing this. In addition, what is the spread of the stress if you refine the mesh around the line? The stress may be singularity because of sharp edge.

Regards,

Ashish Khemka

[Jana Rittelmann]

It might be, I can try refining the mesh more. Thanks for the suggestion. 

But the sharp edge that would cause such a singularity is only there if I cut along the corner, right? So it would not be representative for the full model? And that is why it is better for the uncut corners in the quarter model? 

At least from my understanding, Ansys does solve the model for an eighth of the geometry only (including a sharp edge where I cut the full model to obtain the eighth) and then in post-processing maps the solutions periodically to represent the full model, right?

So would there even be a way to get rid of a possible singularity in the eighth model? Because adding a fillet or something would alter the geometry too much. 

[Ashish Khemka]

It would be difficult to comment if it is singuarity or not. But adding a fillet would give converged stresses and I assume that it would not alter the geometry/ stiffness of the model significantly. The reason to check for singuarity was to make sure if 1/8th model can be used without any issues.

Regards,

Ashish Khemka

[Jana Rittelmann]

Okay, I will check what happens if I refine the mesh and maybe come back with the results. Thanks! 

[Jana Rittelmann]

 

Just to follow up: I did refine the mesh more and the stresses did rise more too. I did not have enough computational power to assess this problem further, so I decided to cut the eighth at a different location outside of the contact area. This helped reduce the stresses and let me focus on the corners by themselves, although there still was some increased stress at the interfaces, I did ignore them for now. 

 

[Ashish Khemka]

Thanks for the update. Yes, this sounds a beeter approach.

Regards,

Ashish Khemka