[GNA] Post-Buckling of arch plate in Optistruct
Altair EmployeeHello
I want to analyse the post buckling behaviour of an arched plate (SHELL Elements) with GNA (Geometric nonlinear analysis, implicit). The model is attached. In another forum (RADIOSS) a similar case was presented, with an pdf file attached, discribing a similar problem :
I am not able to reproduce the results (should be look like this:
I tried different parameters for shell thickness, force size, starting load increment size and solver options (Arc, RIKS,...). Sometimes I get nothing at all, sometimes only results up to point A (s. picture above). How can I get post buckling results?
Thank you
Answers
-
For thickness = 1mm I got to point A in time step 0,2 but not further
0 -
...and how can I plot the time steps on the Y-axis to the Displacements on the X-axis?
0 -
HI,
Are you using OptiStruct as the solver?
0 -
Yes, I choose Optistruct as the solver (I think he uses a coupling with RADIOSS). I managed to get closer to the benchmarke when using FORCE as convergence criteria and as method for LINE SEARCH. However, I cant get the area between B and D. It would be usefull to know how to plot correctly, than I could send you my outcomes at this moment.
0 -
Hi,
Please go through the attached model which is a buckling case solved in OptiStruct.
0 -
Hi,
I have several questions concerning the solver deck you send:
1) No buckling occures? I could increase the load, but anyways, I want to get the deflection of the arched plate from the benchmark (post buckling behaviour including area B-D). The model set up for a general GNA is known.
2) Why using NLOAD in this case? Its static anyways.
3) Why including EIGRL in GNA? It isnt used?
----------------------------------------------------------------------------------------
4) How to plot Load time steps over Displacement (Y-X) in Hypergraph?
0 -
Hi @ahenkes1
I think there is some confusion. The examples are from RADIOSS. Is there a reason for using OptiStruct instead of RADIOSS?
Also, can you point me to the example from which you got the curve?
0 -
As I pointed out, it was just a similar problem. I am used to Optistruct and want to use it, as Optistruct should be capable of solving such a problem (it is a non dynamic behaviour I think, because of that no need for RADIOSS?).
There are different sites presenting this, as it is a rather common benchmark:
https://enterfea.com/problems-with-nonlinear-analysis-convergence-read-this/https://www.comsol.com/blogs/buckling-structures-suddenly-collapse/
https://www.youtube.com/watch?v=JS7dgba2qLc
Similar load path:
And here a different Video with snap-back also:
https://www.youtube.com/watch?v=o5Nu5GF_zZk
----------------------
I want to go through the whole deflection path with the model I attached in the original post. Normaly, the RIKS ARC LENGHT method should be able to get all points .... as pointed out before, with 1mm shell thickness I only get up to point A....
0 -
I only quoted the post from the RADIOSS forum, because it was also a plate, so you can imagine the problem.
0 -
I think I got an update, see files attached. I think it could be the solution, but the post processing isnt working as I imagine:
1) also I solved 1700+ time steps, I only get a view (they are actually ordered by deflection size) to e.g. animate the model behaviour (The time steps solved go ++, than --, than ++ again, as supposed, but the time steps in the drope down menu from HyperView are only from 0 to ++, so around 180).
2) again, how to plot time steps (the actual one solved, not the sorted, censored ones) against displacement? (For validation and comparison)
0 -
Another thing, I was using the geometric nonlinear analysis load step, which indeed uses RADIOSS. I will try the Nonlinear quasi static loadstep with large Displacement and report back.
The post process problem exists anyways, so it would be kind to get help with that.
0 -
Hi @ahenkes1
Please use NLOUT with NLSTAT and also include OLOAD card which outputs applied load vectors.
This will help you to plot the graph for Time vs load.
0 -
Thank you for the tip. I would like to plot the load increments over the Displacement, but maybe with oload it works
0 -
It should work. If you have any problem, do let us know
/emoticons/default_smile.png' srcset='/emoticons/smile@2x.png 2x' title=':)' width='20'>
0 -
Hello,
Know, using NLSTAT, I dont get the expected results. I attached the concerning files. I am not able to plot Load vs Displacement (I cant choose Nodes for either Load or Displacement using h3d, for .res I cant choose Loads as Y reference). Therefore I cant realy say, whether my results are right (as in the Benchmark). The solver doesnt converge at a specific increment, therefore I think he cant go through the whole deflection cuve).
Could you look at the results and try to get a Load vs Displacement graph?
0 -
Hi,
Let us try with non-linear transient instead of NLSTAT.
0 -
Ok, I have no idea how to set it up, could you try it or shell I work into this topic
0 -
'TSTEPNL 7 500 0.0 1 UPW'
*** ERROR # 1000 *** in the input data:
Incorrect data in field # 4.This error I get, no Idea how to go on
0 -
DT is required.
I am trying at my end as well. But so far no luck. I think NLGEOM Impl is somehow giving results
I will let you know if I get any good results at my end.
0 -
Hi, are there any updates concerning our benchmark?
0 -
Hi,
I tired with NL transient. But convergence issues are seen.
I am checking with NLGEOM at my end today
0 -
Thank you !
0 -
Any results with NLGEOM?
0 -
No Success...
0 -
I will update my iteration results soon.
0 -
0
-
Another question: While using 'Nonlinear transient' I want to load 0.25*25 for 0,1 sec and after that linear progression to 1*25 MPa at 1 sec. Therefore I created a TABLED1 with 3 entries:
0.0 -- 0.25; 0.1 --0.25; 1 -- 1. This table I am using for a TLOAD1 table entry and my force with 25 as well. But now my question: will the TSTEP entry in any kind overwrite this? If I have N=100 and DT=0.01 will the solver still use the table entry?
0 -
I think I solved it for my own, of course the N,DT combination just divides the given load profile in increments
0 -
Hi @ahenkes1
Sorry that I couldn't able to reply due to long weekend
At my end Transient NL run fails with convergence issues at exactly 95% load. I would be interested in looking at your model
0