File will not calculate!!! Lack of convergence of nonlinear analysis

[Engineer3636]

Can someone please help me. I am trying to calculate the file in the attached link but it it wont calculate. I usually check structural verification at this stage and it shows me an error. But this time it says nothing is wrong. Please help this is urgent.

Thanks

https://www.wetransfer.com/downloads/740a3001040c612691fc56519d48408b20130821101251/c7b341

[Artur.Kosakowski]

Try:

If you find your post answered press the Accept as Solution button please. This will help other users to find solutions much faster. Thank you.

[Engineer3636]

Arthur,

That's brilliant works fine. But when should I be changing Photo 1? Also regarding photo 2 does it affect the result if I select continue?

Thanks again

[Artur.Kosakowski]

I'm not sure if I understand you correctly but in general when analysis diverges the results are incorrect. The 1e-4 precision is a very strict limit so you can increase it when you see that the solver has got difficulty to reach that level.

If you find your post answered press the Accept as Solution button please. This will help other users to find solutions much faster. Thank you.

[Engineer3636]

Hi Arthur, I have another panel that will not calculate. I have tried to change it like the previous one it still doesn't work at 0.005. Should I change this again? Any advise would be much appreciated.

Link below

https://www.wetransfer.com/downloads/41eb5de9c15c65f861dacc50b2920a5a20130829170818/235e9a

[Artur.Kosakowski]

Try to apply load in one increment

If you find your post answered press the Accept as Solution button please. This will help other users to find solutions much faster. Thank you.

[Engineer3636]

Ok Artur, this has allowed me to run the analysis but what affect will this have on my results. As you know I am recently new to Robot. I never usually change any of the preferences like this. I usually only tick the box which states "matrix update after each iteration" and it works.

Thanks

[Artur.Kosakowski]

The effect will be that the analysis converges which is the must for the correctness of the results 

For models that are difficult to converge it may be necessary to adjust some parameters of the analysis to obtain the solution. Sometimes applying load in one increment helps but sometimes it is better to apply it in large number of small increments. There is no general rule that I could write - you just need to try and  check what happens.

If you find your post answered press the Accept as Solution button please. This will help other users to find solutions much faster. Thank you.

[Engineer3636]

Thanks Artur,

Would it be wrong then to change the value from 0.005 as previous did to say 0.01. This is obvisouly less strict. Is 0.005 the most you would change this by?

Regards

Daniel

[Artur.Kosakowski]

It is really hard for me to tell what value is still good enough and which is wrong. Personally I do not want to go over 0.005 and from my perspective if that does not help it is usually something 'wrong' with a model itself. You may try to set 0.01 as a sort of 'desperate' action but then the extra care should be done for checking displacements and rotations as well as sum of reactions vs. applied forces and values of residual forces.

Please note that for different kind of convergence issues from this forum no one (if my memory serves me well) has actually been solved by setting the tolerance as large 0.01.

If you find your post answered press the Accept as Solution button please. This will help other users to find solutions much faster. Thank you.

[Engineer3636]

Hi Artur,

Another panel wont convergence. Any ideas really need this to work. See below

Thanks

https://www.wetransfer.com/downloads/8a52ed89c5c8968a4287c14e123a7fb720130924121406/bd43c8

[Pawel.Pulak]

Your model in not convergent even without the 2nd and 3rd order geometrical effects (non-linear and p-delta check-boxes), when leaving only the nonlinearity of compression-only bars

First 2 reasons of convergence problems:

1/ model is unstable - see the screen capture below

It can be noticed that the structure is free to rotate:

- about the vertical axis passing through all collinear supports

- about the horizontal axis parallel to global Y and passing through the top support

Another story related to it, why these instabilities were not reported by the solver, is to be investigated:)

2/ leaving incoherent mesh and using kinematic constraints deteriorates convergence

As concerns instability of the model it should be solved by you using correct support conditions, adequate to the reality - I do not know it. Temporarily I have solved it adding the support fixing UX and UY in bottom right corner of the structure.

As concerns kinematic constraints, I have switched them off in job preferences, activated the iterative adjustement of the FE mesh, I have unfrozen meshes, regenerated them and frozen again.

As concerns parameters of analysis:

- I have left only the non-linearity of compression-only bars

- set BFGS method with 150 iterations (the tolerance left as 0.0001)

- set cases 2to4 as auxiliary ones because convergence is really necessary for combinations 5to8

After above modifications the analysis of cases 1 5to8 is convergent.

Such model attached - without results to reduce the size.

Of course if you apply the support conditions corresponding to the reality it may be different...

Regards,

[Engineer3636]

Thanks Pawel,
It appears my restraints for the opposite side was deleted. (don't know how) - It still wasn't working it was correct. I will amend and see what happens
Regarding the switching off constraints and activated the iterative FE mesh - can you show me how this is done.
I am still trying to learn robot
Thanks

[Pawel.Pulak]

See the screen capture below:

It was mentioned by Rafal and Artur several times in other posts.

Regards,