Difficulties with nonlinear static simulation

[Anonymous]

Hello,

I have spent slightly more than a week trying to run a nonlinear plastic static simulation in Fusion360.

The linear static analysis shows no problems when we run it.

I have already:

- simplified the geometry

- improve the mesh at locations with high tension

- modify the material strain-stress curve

- consider an elasto-plastic (bi-linear) material

- touch some parameters:

          (general) number of steps -> from 10 to 999

          (mesh) Model-based Size -> all the range

- reduce the loads in order to ensure that the structure can withstand them

- modify the constraints, from small areas to larger ones, and from friction-less to fixed

I still run into errors everytime I launch a simulation. Here some exemplary warnings and errors I get:

- Warning: TET ELEMENT 97002 HAS AN EDGE POINT RATIO GREATER THAN 0.5

- Warning: TET ELEMENT 249327 HAS SINGULAR GEOMETRY

- Warning: STIFFNESS MATRIX SINGULAR OR NON-POSITIVE DEFINITE

- Warning: DIFFERENTIAL STIFFNESS MATRIX IGNORED FOR THIS ITERATION

- Warning: STIFFNESS MATRIX SINGULAR OR NON-POSITIVE DEFINITE

- Warning: BISECTING CURRENT LOAD INCREMENT

- Error: Local solve module failure, Error: 1
- Error: Solver Error
- Error: An error occurred while solving the model.

At this point I have not clue about how to move forward. However, due to he warnings I get, I guess the difficulties lie in the mesh.

At least, I would like to know where in the model are the elements that are reported in the warnings, so I can focus on improving the mesh at those areas. I have though on generating a *.vtu results file from the TEXT COMMANDS window with the SimResults.ExportActiveResults command and check it. I can only do it after getting a result, which I get only for a linear analysis. Two issues I have found with this approach:

- I need to create different meshes for the linear and for the nonlinear analysis. Therefore, they may not be identical and the element numbering may differ completely.

- The .vtu results file does not include geometry details. I can only see stress, connectivity and element type details.

Could someone provide me with a hint about how to move forward?

Many thanks in advance,

A.

[John_Holtz]

Hi @Anonymous 

The singularity errors will stop the analysis. Are these occurring at the beginning of the analysis? This would indicate a distorted mesh. Or is the singularity error after some  number of steps? This would indicate that the element is distorting due to the stress/strain. (It has been a while since I ran a nonlinear analysis, so I am not sure if Fusion provides any feedback to know whether it is occurring at 0% complete versus 50% complete version 99% complete.)

Fusion does not have any tools to identify specific elements (or nodes). When such issues occur, we use internal development tools to look at the actual analysis files and locate the specific elements. Can you export the model and attach the .F3D file? (I received your private email with a link, but I could not see how to download the model from the link.)

[Anonymous]

Dear John

Thank you for the reply.

Good to know that the problems could be related to the mesh, either because it is poorly generated (scenario A) or as a consequence of distortion under applied loads (scenario B).

Considering that the time that goes by before the analysis stops is considerably larger than that required by a linear static analysis, I would say that we are in scenario B. However, we do not have any feedback about the steps that the solver has already run... 

On the other hand, we have already reduced greatly the loads applied to the structure, so that it should not undergo large deformations. This is not coherent with the previous conclusion.

Even in scenario B, we plan to improve the mesh in order to increase our chances to get a result. Reviewing the mesh control parameters available, we have some doubts about:

- (check button) "Scale mesh size per part". I guess it does not apply to our case, as we are working with a single part

- (check button) "Create curved mesh elements". What would be better ??

- (slider from small to large) "Max aspect ratio". What are the numeric values at the ends of this slider ?  We plan to set at at about 25%, close to the small end.

I wonder about the possibility to run some kind of mesh quality check ? When we run an analysis, we already get reports about elements with conflicting geometry (EDGE POINT RATIO GREATER THAN  0.5 and SINGULAR GEOMETRY errors). Is there some other way to perform a more detailed analysis of the mesh ?

We are sharing privately the .F3D model file, both with a linear and a non-linear analysis defined. (Loads are much smaller in the non-linear one).

Regards,

A.