Duration of nonlinear analysis

[Anonymous]

I am trying to perform nonlinear static analysis for my model. I am trying to get new stresses and displacement for  my 100+m building.

Right now it has calculated for 24h and it seems that it does not want to stop.

Is there some way that i can still achieve the results with getting smaller accuracy or set other options to speed it up.

I have set the nonlinear part only for one combination,added the current status and options used.

I cannot figure out, interation number means that it will do maximum 40 of them in one calculation? and then if it is successful before 40, it moves up to the next increment?

By that logic if it does not fail, it will take a week? 

Any wisdom or help is welcome

[teixeiranh]

File please!

[Anonymous]

It finally calculated it and the result is "Not convergence of nonlinear problem. Case number=20. Report of the current file will be written to the log file."

As i understand then it could not perform the anaylsis. What can I do with it now?

One idea is that It does not like the pile foundation which is modelled as  elastic springs?

The results gained with the failed anaylsis are unrealistic as the deformation is 10times smaller than it used to be.

What can i do with my file to improve the situation?

[Artur.Kosakowski]

Do you have the instability of type 2 reported for linear static too? If yes, than you should address it first.  My assumption is that you have unidirectional supports which allow for uplift. If so check the overturning vs. stabilizing moments. You may try to decrease horizontal loads and check if the analysis converges. 

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.

[Anonymous]

Instabilities are coming from the fact that columns are changed into pinned-pinned releases. Ideas was to take the horizontal load with walls, not with columns.

It has been there for ages, even when there were no piles and elastic supports (everything was fixed).

Could you please exaplin more about  "check the overturning vs. stabilizing moments"

[Artur.Kosakowski]

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.

[Anonymous]

All my supports are elastic springs, never used unidirectional supports.  The question is how can i actaully check the push-over of the WHOLE building. 

In fact I know that some of the supports i am using are going under tension but thats less than 10% of the supports and the building should not be that unstable. (just the first row of piles does that)

http://www.4shared.com/rar/XDyjqmqWba/Tornimae__vaiadega_150425.html

Would it help if i removed the springs and piles and replaced them with fixed supports?

[Anonymous]

Found out that much. When I change the end support of the pile it calculates the results. 

So It does not matter what is the horizontal deformation. It does not like the vertical elastic support. Did i understand correctly, that the nonlinear anaylsis does not like when there is elastic support in tension?

[Artur.Kosakowski]

Could you post the fine on another server please? I have some issue to download it from 4shared.com.

[Anonymous]

https://www.filemail.com/t/ffbe60bad7de49ad999d7a9de9f69129

Things have moved on a bit. When i removed the piles on elastic springs, then the matrix is negative. That should mean the building is unstable at such loads?  So i need to make my bulding stiffer or there could be some other reason?

When i use the same piles on smaller, stable building then it still is unable to perform unlinear analysis, it lasts forever and there is nothing is done.

[Artur.Kosakowski]

Things have moved on a bit. When i removed the piles on elastic springs, then the matrix is negative. That should mean the building is unstable at such loads?  

Correct.

So i need to make my bulding stiffer or there could be some other reason?

When i use the same piles on smaller, stable building then it still is unable to perform unlinear analysis, it lasts forever and there is nothing is done.

I wanted to check what is the buckling coefficient for linear buckling analysis for combination you marked as nonlinear therefore I set all cases to be linear static and I set buckling analysis for  the selected combination. I ignored incoherent meshes and instability but this resulted in error 3060 during the buckling analysis (which is probably the same reason why nonlinear analysis gets into the stage when it runs "flat" with no end on the horizon). This leads to the following conclusion:

The reported instabilities cannot be ignored meaning you need to check what causes them and solve that before moving on (perhaps you need to look at the meshing first?).

[Anonymous]

I manage to solve the stability issues, It seems quite impossible to make the mesh perfect as  by eye all the nodes are connected and nice but still says those spots are incoherent.  

By solving the instability i manage to get the buckling modes 1-10. added outtake from it. 

but STILL I am not able to perform nonlinear analysis. I let it run over 12 hours and it could not do even one interration, which for me means it will pretty def. fail. Just to make it clear, I did all this on my model which does not have piles. All piles are replaced with regular supports. 

Just in case there is something more, 

https://www.filemail.com/t/ddf0f7c49cf4481892791d3cb2e21110

I really appriciate your help

In addition, can you explain why are the piles on elastic springs making nonlinear analysis fail. Without piles on regular supports it gives realistic results and when I added the piles, it fails to give any results at all. Is it possible to perform such anaylsis on such piles (elastic springs in hor and vert direction)at all? You can just give yes or no answer. 

[Artur.Kosakowski]

I'm not sure if I understand the current situation correctly. I assume that this is the model you corrected 

https://www.filemail.com/t/ddf0f7c49cf4481892791d3cb2e21110

and if I set nonlinear analysis (which in RSA 2015 means P-Delta) for combination 20:ULS17 it converges ( I have used Robot 2015 SP5 for calculations). The next step should be to define piles with elastic supports for this model. Assuming that the provide boundary conditions that prevent large displacements of the building (you may check this first with linear static analysis) the nonlinear analysis should converge too. 

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.

[Anonymous]

For my suprise  the version without piles convergences. When doing the same thing on the version WITH piles, it  goes a bit wrong.

I have set the convergence limit for 5.0E-003. It does it until 3/5 and then it starts to stay gather in 7.0E-003 line.  When i reach to 4/5 with increment number then it stays stuck at  1.11E-0,02. (starts getting almost the same value for 30times in a row)

Up to what level can I set the limit to? When i read the previous posts then the 5E-003 seemed reasonable but when setting it to 1.2E-002

As i understand the accuracy decreases when I change the number higher but..?  how far in your oppinon could i go, I imagine that increment number 5/5 requires convergence 1,3E-002. 

https://www.filemail.com/d/eedftbsgtfzpnev

Added the file just in case. 

[Anonymous]

My mistake, It does not convergent. I used Robot 2016 for my last calulcations and i noticed uptill now that you switched the names. Pdelta according to 2015 does not still work.

And also the last file, its not up to date, the "ULS13" should be nonlinear, everything else is correct. 

[Artur.Kosakowski]

As far as I can see you just ignored reported instability warnings. If you checked obtained results for linear static analysis you would have noticed very large RZ rotations. There is no way second order analysis converges for such values 

I have added RZ restrains to your definitions of supports which solves the instability and allows the nonlinear calculations to converge. 

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.

[Anonymous]

You mentioned you added rotational restrictions to the support? which support do you mean and do you mean just for that node or for all the nodes?

[Artur.Kosakowski]

For my test I blocked RZ in all supports definitions from your model. 

[Anonymous]

It must be me.... I changed all the elastic supports ( blocked them in RZ). Got rid of the instability but the analysis will not go futher than 2/5.

[Artur.Kosakowski]

I'll send you my model (I used RSA 2016) so that you can check the parameters of nonlinear analysis I set as well as the definition of supports which you may play with if necessary.