Oscillating behaviour - No convergence of non-linear problem

[Anonymous]

Hello,

In my model I recently changed some elements from simple bars to compression or traction truss bars, and this change led me to the classical "no convergence of non linear problem". It is a problem that I had already solved in previous analysis of the same model, following the guidelines found on these forums. But it seems nothing can help after this last change.

More precisely, the non convergence concerns only one particular wind case, and although I tried modifying extensively all the recommended parameters, I always get at some point an oscillating curve during the calculation, as you can see in the screen capture underneath. I must say I don't precisely understand what this curve represent, but I was hoping someone would know more about such phenomenon. I must also add that I already had truss elements in my model, but it was still functionning. 

Thanks in advance!

Aubin Cortale

PS: Attached are a capture showing the oscillating phenomenon, and pictures of the parameters set for the analysis (the current parameters, as I already have played with them quite a bit, in the limits of my knowledge).

[t.sautierr]

Even if it not really a satisfying solution as it doens't solve the problem : if this happens for a wind-only load case, it is can be ignored as this a load case that does not happens alone -> set this load case to auxialiary and focus on the combinations.

Otherwise, does it happen for the wind in this opposite direction ? ( -wind 2) does it happens if you try with half the load (actually before is start oscillating around a precise step of iteration? Is you structure symetrical? are the tension/compression bars also symetrically distributed within the structure?

[Anonymous]

Thanks for the answer, or should I say the questions! So to answer them:

- I just tried adding a 0,5.WIND_2 combination, with the same parameters (to act as a load reduction as you said), while letting the previous WIND_2 run also. As you pointed it, I tried to see at which point the oscillation was beggining, and it seems to appear always at the last increment. Strange thing is that this time WIND_2 could calculate! (I didn't change any parameter nor any geometry, only added the 0,5.WIND_2). I tried it twice now, and the second try confirms the success of the first, but lets me completely confused. The oscillations still appear, but Robot finally solves it after a few other increment which come if I'm not wrong from the "Increment length reduction" parameters. So I guess my problem is solved somehow (and all the combinations are working), but I don't get how, so I'm not exactly satisfied.

- Stranger yet, 0,5.WIND_2 had exactly the same curve behaviour (with oscillation showing at the very last increment, and Robot trying to add other increments) this time without success. I'm reaching a level 2 confusion. 

- I did the -WIND_2 combination, which worked like a charm. This is not a symetrical case though, if you had that in mind.

- For the other questions, yes my structure is completely symetrical, compression/tension bars and loads included. 

I guess I could just be happy the case I wanted finally worked, but this 0,5.WIND_2 is really strange.I would love to attach my file, but it is pretty heavy,around 500Mo, so I guess I'll send it in another post.

Thanks again for your time.

[Artur.Kosakowski · Accepted solution]

As t.sautier suggested in case of model with tension only bars you may think about the design situations that actually do happen in reality and use the other load scenarios as auxiliary ones that you need only to build load combinations. In this case there is no chance that the wind will act alone without the self weight of the structure so if the wind load case converges or not is in fact irrelevant. For side loads it may happen that if you apply certain amount of load (load increment) there is compression in tension only elements due to the structure deformation so they are switched off but when you increase the load (next load increment) the deformation changes and some of the bracing that were prebviously switched off are now under tension so they work again. This path repeats itself with each step so you see at the "oscillation" and the analysis finally diverges. I hope this makes the situation you observed easier to understand. 

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]

Oh thanks I see, now I understand more clearly the situation. And given the model I can be almost sure what you said is precisely what is happening. Many thanks for the replies to both of you. 

Aubin Cortale