algorithmic modeling for Rhino
I am experimenting with Karamba module for the first time and want to optimize my truss. It's made of GLT and I have problem with the specification of "fy" value, which influences the result a lot. Can you please provide me some relevant information how to set up material that behaves close to characteristics of GLT?
I am getting some crazy outputs now...
See attached files.
for GLT a characteristic strength of roughly 1kN/cm2 is OK. The exact value depends on the chosen product.
The main reason for the strange results was that you selected a rectangular hollow cross section with a default wall thicknes of 4cm. In the attached definition that is changed to a full cross section.
The second problem was the buckling length. Karamba assumes it to be the distance between the endpoints of the beam (see the manual for details). This is too large in your case. The correct value depends on the Eulercase of your column. If both ends of the column are fixed against translations a value equal to the column height would be correct.
Oh that is a big difference now!
Thank you so much for the correction.
I have a suggestion, if it's not better to feed Karamba only with units using [m] or [mm] since the majority of tables of material characteristics are using those. Sometimes [cm] is a bit confusing. Or the best would be the option to set it up (sorry if the option is already there..)
After I bought Karamba, I tried optimalization of cross sections of my Glulam structure again. Later I did the evaluation by hand and got way different results. For example: Karamba gave 1,5 meters high truss with width 0,3 meters. But the evaluation I did by hand told me, that for this height I need just only 0,15 meters width element.
Can you please see my file, if there is some sort of mistake? Will there be any possibility to work with better material specification soon?
the cross section design was governed by the displacement limit you set.
Another problem was that the incoming data of the 'Assemble'- and 'Otimize Cross Section'-component were not flattened resulting in multiple models.
I changed this and a few other things. Please see the attached, annotated definition.
Hello again. Thanks for your help. I appreciate that a lot. I am glad I've bought the software and also that you guys offer such a perfect support.
This time I changed the structure and did Karamba optimalization again. It's workig pretty well, I went through the manual as well. But there are some questions, some strange behavior of elements. See the attached file and picture showing the problem.
1) elements should be loaded symetrically, but it seems that one direction is more significant than the other. I was thinking if it could be because of joint stiffness? Is there a way to tell Karamba, that all the joints are stiff? (a piece of metal connecting plates together with screws..) Or what could it be? Different algorithm?
2) I am getting some overlapping elements, which could be because of some overlapping elements, I guess. But that's not such a problem, as the first one.
try to change the value of the input-plug 'ULSIter' which can be found under 'Settings' of the 'OptiCroSec'-component. It determines the number of cycles used for updating the cross sections. You will get a uniform distribution of cross sections for e.g. 'ULSIter'=1. Higher values (e.g. 30) lead to a more statically determinate system where primary and secondary structural systems emerge.
Hi Clemens. Thanks for the hint, I did it that way. But the two directions are still much different than expected. Look at the picture attached: one direction is much bigger than the other and I don't know why. I saw some tutorials from Karamba and it might be because of joints, but when I tried to add them to the model, I got huuuuge elements :D so maybe there were mistakes.
The grid is made of many short pieces of wood connected with solid steel joint together, so the structure should be rigid.
try to set 'ULSIter' and 'DispIter' to '1'. The 'ULSIter' input determines the maximum number of iterations for the Ultimate Limit State. 'DispIter' controls the number of iterations used for satisfying the displacement limit.
The cross sections are still not the same but I think that comes from the slightly irregular geometry.