Quadratic Tet vs Linear Tet

[Anonymous]

Hello,

I've been doing some testing on different element types and sizes on a small rotor to determine bending stresses caused by the Poisson effect. Initially, I noticed that using linear tetrathedral elements does not model the bending at the trunnion support very well but the axial stress in the centre of the rotor execellently. I have tried to accuratly capture the bending stress using quadratic elements however the stress field, as you can see, is very erratic.

Is there a way to get a more consistant result using Quad elements? I am trying to make design modifications at the trunnion to reduce this effect as much as possible.

Thanks,

Cody

PS - I am not using brick elements since I consistantly have water tight issues.

[Anonymous]

Sorry, it might help to note that the stresses are caused by the roatation of the rotor and that there is a support at the top bearing surface restricting movement in the x-y-z direction.

[S.LI]

Usually, the second order tet is better than the  linear one, especially for bending problem.

Just have several questions:

1.) linear static or MES?

2.) did you have stress output on mid-side nodes?

3.) what is the your package version?

BTW: I didn't see the obvious difference between the two pictures. Maybe you can highlight it for me.

Thanks.

[Anonymous]

Hi:

Here are the responses for the questions:

1) Linear

2) I didn't realize you had to "turn it on". It seems there are a lot of steps to goign from linear to Quadratic. You have to include 2nd order mesh types, make your element include midside nodes, and turn on mid side nodes in the output?

3) Version 2012.01.00.0017-W64/X64 15-Jun-2011

Sorry I meant to only include one of the pictures but it seems they both were - they are the output for the Quad run I did.

I've tried running it with mid-side nodes in the output file as well as added what I get when I run it with Linear Tets.

Please have a look and let me know what you think the problem may be.

Thanks,

Cody

[sunj.autodesk]

There is a fix in Ranger (2013) release related to stress results in tet elements with middle side nodes. Can you download the 2013 Beta to verify whether your issue is fixed?

If you post your model (if it is not too large) here, I can also check it for you.

[Anonymous]

Where can I download the Beta?

[John_Holtz]

Sorry Cody, but the beta software is no longer available. The final version of the software is scheduled to release in a few weeks. (That's always subject to change.)

I see that Jun offered to run the analysis here. The page "Create, Post, or Provide an Archive of your model" indicates how to create an archive.

To return the results to you will probably require an FTP site or use of a service like "You Send It". I will contact you through the "private messages" to get your email address.

[Anonymous]

Please see attached for my model. I've turned the resolution down so that the file will be smaller.

Thanks

[sunj.autodesk · Accepted solution]

Hi Cody,

I noticed you are using Brick element with the model meshed with all tets. In this senario, brick element will be degenerated to match the tet mesh, and make result precision questionable.

You can simply change the brick element to tet element. After doing that, the results looks all good on my computers.

Let me know if you still have problems for your model.

Thanks,

[Anonymous]

Changing to a tetrahedral element type helped smooth the results out significantly. It was a little counter-intuitive since Brick & Tetrahedral meshing options are available in the pure brick element type.

Is it possible to model this in 2D using Multiphysics since the stress feild should not change with respect to Theta since this is simply circular rotating part? This would make identifying stress concentrations much easier.

Thanks,

Cody