Discussion Groups

Simulation Mechanical and Multiphysics

Reply
Contributor
JohnBeach
Posts: 21
Registered: ‎05-24-2012
Accepted Solution

non-linear residual of 1.00

566 Views, 9 Replies
01-17-2013 08:21 PM

I have a model that runs in static linear. It will not run in static non-linear. It simply runs through the 19 time steps and constantly returns a residual convergence of 1.

Can a developer or Autodesk person have a look at it and see what I have overlooked? I have successfully changed the analysis type on a smaller model before.

 

I copied across to a new design scenario.

changed all plates to shells

changed element definitions

re-assigned material to be sure

re-assigned pressure load

re-assigned constraints

check analysis parameters and added a load curve

 

I tried to attach and ach file; but it was 11 MB. To whom would I send that file?

 

Assistance appreciated,


John

Please use plain text.
Valued Mentor
AstroJohnPE
Posts: 501
Registered: ‎08-30-2012

Re: non-linear residual of 1.00

01-23-2013 12:40 PM in reply to: JohnBeach

Hi John,

 

One suggestion is to set the load curve(s) magnitudes to 0 so that there is no load on the first time step.

  • If this solves, then it may be that the initial load is too large.
  • If this does not solve, my guess is that something in the model is statically unstable.
  1. Are all of the parts connected node-to-node? (Linear stress has "smart bonding' which connects parts even if the nodes do not match; MES does not have this capability.)
  2. The only other thought that comes to mind is the solver that is used. Either the log file or summary file (accessible from the Report tab) indicate which solver was used, such as the iterative solver or sparse solver.

Otherwise, did you try the tips in this post to reduce the size of the archive? http://forums.autodesk.com/t5/Autodesk-Simulation-Mechanical/Create-Post-or-Provide-an-Archive-of-yo... Maybe you can get is small enough to post.

John Holtz, PE
Mechanical Engineer
Pittsburgh, PA

16 years experience with Simulation Mechanical
Please use plain text.
Employee
zhuangs
Posts: 216
Registered: ‎10-04-2010

Re: non-linear residual of 1.00

01-29-2013 12:00 PM in reply to: JohnBeach

Hi John,

 

You can send the model to me: Shoubing.Zhuang@autodesk.com, better via Accellion

 

I can help take a look at the model.

 

-Shoubing

Please use plain text.
Contributor
JohnBeach
Posts: 21
Registered: ‎05-24-2012

Re: non-linear residual of 1.00

06-11-2013 05:09 PM in reply to: zhuangs

Thanks Shoubing. I have emailed file.

 

Also here is link to ach which I think works.

 

https://dl.dropboxusercontent.com/u/42633891/1926-CA-0211%20A.ach

Please use plain text.
Employee
zhuangs
Posts: 216
Registered: ‎10-04-2010

Re: non-linear residual of 1.00

06-12-2013 07:59 AM in reply to: JohnBeach

Hi John,

 

If you can change the "Element Formulation" in "Advanced" Tab of Elment Definition of the 5 Parts (1~5) from "General" to "Thin", the model will run to completion smoothly.

The 5 Parts (1~5) have the other dimensions much larger than the thickness, so "Thin" shell will be appropriate for them.

 

-Shoubing

 

Please use plain text.
Contributor
JohnBeach
Posts: 21
Registered: ‎05-24-2012

Re: non-linear residual of 1.00

06-12-2013 04:03 PM in reply to: zhuangs

Thank you for suggestions.

Manually setting the solver to sparse (it was automatically going to iterative AMG) allowed it to solve in 20 time steps and 1.5 hours.

 

Setting the element formulation to thin allowed it to run; but at 25% started reducing time steps to solve so I stopped it.

 

I guess I will try thin and sparse.

 

I would have thought it would automaticallly try sparse if iterative did not work. I would appreciate comments.

Please use plain text.
Contributor
JohnBeach
Posts: 21
Registered: ‎05-24-2012

Re: non-linear residual of 1.00

06-13-2013 01:35 PM in reply to: JohnBeach

The thin and sparse idea made it to tie step 9 of 20. IS this a problem with the thin elements perhaps?

Please use plain text.
Employee
zhuangs
Posts: 216
Registered: ‎10-04-2010

Re: non-linear residual of 1.00

06-17-2013 11:46 AM in reply to: JohnBeach

Within their capabilities, "Sparse" solver and "AMG" solver should provide similar results.  However, "Sparse" solver usually works for small or medium models, while "AMG" solver works for large model.

Once "Automatic" option is defined, the code will automatically select one of them based on the size of equation.

 

-Shoubing

Please use plain text.
Contributor
JohnBeach
Posts: 21
Registered: ‎05-24-2012

Re: non-linear residual of 1.00

10-14-2013 06:34 PM in reply to: zhuangs

I have this problem again. I am applying a hyrdostatic pressure and it looks like the solver does not recognize the pressure. I put a simple force on it and it does run.

Please use plain text.
Employee
zhuangs
Posts: 216
Registered: ‎10-04-2010

Re: non-linear residual of 1.00

10-15-2013 06:44 AM in reply to: JohnBeach

Hi John,

 

Copying a design scenario (Linear) to another one (Nonlinear), we do need to click the "Element Definition" for all the parts, since Nonlinear and Linear use different AGSDB variables, defined by UI.  You did right.

 

To use "Static Nonlinear with Nonlinear Material Models" analysis type, the model must be static stable and well constrained to avoid any rigid body motion.  For your model with shells, the model should not move along x-, y-, and z- directions, and not rotate along Rx-, Ry, and Rz.  Otherwise, the model generally cannot converge with "Static Nonlinear with Nonlinear Material Models" analysis type.

 

If you think the model is well setup, you can try "MES with Nonlinear Material Model" and see this can help.

 

You can send the model to me via email.

 

-Shoubing

Please use plain text.