Type 2 instability and no convergence problem

[danker1988]

Hello,

I have a strange shape structure that I want to analyze with Robot and I am not able to correct some warrnings and an error. Can somebody have a look at the model and suggest something to fix it?

The warnings are about type 2 instability in the Rz and Ry direction, and the error is for no convergence of nonlinear problem (I am currently analyzing just the case of the self-weight of the structure).

The Robot model file can be downloaded from the following link (it is to large to be attached here).:

https://dl.dropboxusercontent.com/u/54052628/split_lines_with_cables.rtd

Thank you for the help,

Dan

[t.sautierr]

When you define cables, you need to set the installation parameters : force/stress at the beginning or lenght/strain. Beware the there is a fundamental difference between setting one or another (see RSA help).

Your instabilities are coming from not defining these parameters (as your structure is stable with no cables). For example set the elongation to 0.01 (cable unloaded) and the problem will converge.

in the other hand, when I've tried to set -0.05 (cable in tension at the beginning), the problem wasn't converging. I assume is due to a shape effect (spiraled structure). To load the cable at the beginning, you shall do it for part of it I assume (set negative elongation for some and positive for others). Then to simulate post-tensonning -> temperature.

A another thing : do you really need to have a quarter of your bar below 0.05m ? such an amount of bars makes the calculations longer, maybe 0.1m is enough?

[danker1988]

Thank you for your answer.

I might have understood wrong the help documentation regarding cables in Robot. It was said that "Stress, Force, Length and Elongation are mutually exclusive and are not required. If you do not specify any of these parameters, the length of an unloaded cable equals the distance between the nodes.", therefore I understood that if I do not enter these parameters, Robot will take automatically the distance between the nodes as the length of the cable (the L parameter). I will change the definition, and see what I obtain. 

Regarding the small length of the bars, it is not mandatory. I would like to simulate as good as possible the actual shape of the structure. I have a DWG file with the center lines of the profiles, and I have imported that into Robot, but I could not optimize the import process. I know that the current way the structure is composed will make the analysis slow, and that it will be also difficult to apply the loads on it, so if you have any suggestions on reducing the number of bars when importing, please let me know (I have tried different combinations for the arc discretization parameters, both with constant number and number of full angle divisions, but the number of bars created was the same). 

[t.sautierr]

Yes you cold have let it undefined but usually it leads to non convergence, it is better to give some tension even if very small than nothing. RSA meets difficulties to solve the problem if nothing is defined.

Regarding dwg, you can divide your arches in autocad, divide it to obtain the lenght your want and then draw polylines joining these points, then import it into autocad (it avoids to have the discretization in Robot).

Otherwise send your dwg and I'll try to discretize directly into Robot.

[danker1988]

I have considered dividing the structure before importing it into Robot, and I have already done that to a certain level (the initial structure was only made out of 45 curves). But I would like to avoid this; I already analyzed two similar structures and I will have even more complex ones, so I would like to know how to better control the dwg/dxf import into Robot. Can you explain how to do this, or show me where I can find more relevant documentation on the subject? Also, if different file formats work better, please let me know; the original structures are usually in Rhino and can be converted to different formats. 

[t.sautierr]

dwg is fine. please upload your dwg and I'll try to discretize it in a more simple way, I 'll make a video on your example if I succeeded.

[t.sautierr]

In you dwg, the polylines are constructed with many points, that's why you get all these bars when importing to Robot.

During the importation you can divide element in more little one but you can merge them if you have too precise drawing.

So method 1 : redraw the arches (bottom part, because the part with cable is fine, one element between two cables) with less points (for example draw 1 arc approaching your polylines).

method 2 : redraw the arc (bottom part) in RSA (object, arc) and specify that you want to divide it in XX pieces and that RSA creates bar for each pieces