Optimizer Beam Karamba + Galapagos

Replies to This Discussion

Permalink Reply by Nicholas Rawitscher on November 20, 2016 at 11:22am

Hi Alexia,

Well first of all what is your "fitness function" goal, i.e what are you trying to reduce or maximize? is it just minimizing the displacement of your structure? Why is galapagos only controling the cross section`s dimensions and not also the parameters of your shell?. My experience with galapagos  is that the most important part is coming up with a proper fitness function which is basically a mathematical formula in which you inform the solver what you want it to search for and the optimal target for the fittest solution. 

Coming up with a good fitness function is often a challenge and it needs to be mathematically correct in order for you to have interesting and relatively precise results, otherwise it will not do what you expect it do do. Fitness functions become more critical as your model is more complex. 

I suggest that you start digging  in to genetic algorithms so you can also re-evaluate your search if you need to, and understand the idea more in depth. also Octoupus is another good evolutionary solver which focuses on multi-objective search and you can determine inside the component the min and max values of your fitness function. In this case it can be better to use Octopus over Galapagos if you have more that one fitness function as a workaround in coming up with a proper mathematical formula.

I hope this helped a little,

good luck with your work

Nicholas

Permalink Reply by Alexia Ch on November 20, 2016 at 11:41am

My intention is not to redimension the shel but only to find the most appropriate beam dimension with the minimum displacement. But not of the order of millimiters as it does but in lets say 20cm of displacement. more or less.

Permalink Reply by Rafael Pastrana on November 20, 2016 at 7:30pm

Hi Alexia,

One thing that you could do would be to use the CrossSection Optimizer algorithm in Karamba. 

Input a list of cross-sections from which the algorithm can look for in order to perform the structural optimization according to Eurocode 3, and define the expected level of utilization you would require.

Fuerthermore, you can input a maximum deformation value (in meters) too so that the algorithm takes it into account.

Find attached a reworked file.

Hopes this helps,

Rafael

Attachments:

• Voronax400ptconfrontoperkaramba-rp.gh, 46 KB

Permalink Reply by Alexia Ch on November 21, 2016 at 11:45am

yes my intention is to use both also the optimizer of Karamba and the galapagos. to see aftewards in Sap2000 which is the best solution. But I can't make it work :( :( :( since this is the simpliest example of my thesis and I can't make it work.... I'm so stressed out, since I loved Karamba and the paramatric world I don't want to surrender! thank you so much for your answer. can I make you a couple of questions?? at the cross section selector at the voice Name|id you put 0 y? also the same at the optimizer the max util u put 0.6 what that supposed to be? and the last question y at the supports you put the tx tz ty only and not all? my prov told me to impose all of them also the rotations at the base of the structure.

Thank you so much for the answer and for the file. you saved my life! :P

Permalink Reply by Rafael Pastrana on November 22, 2016 at 7:54am

Hi Alexia,

It's great to know that the file worked! 

Regarding your questions, at the Name|Id (which means Name or Index) I input a zero so that the component knows that it should take the first Cross Section and assign it to my elements as my first version for design.

At the Optimizer, the 0.60 means "try to make my elements work at most at 60% (out of 100%) level of utilization".

Well, in practice it's more realistic to impose "pinned" supports (only displacements are restrained) because otherwise your foundations would need to be design to resist bending moments, and there is not such a thing as a perfectly "rigid" support (rotations and displacements fixed).

However, you can check both options and evaluate at the outcomes by yourself. That's part of the fun in Karamba!

Best,

Rafael

Permalink Reply by Alexia Ch on November 21, 2016 at 3:48pm

and the last question, I've seen that you put a mesh load of 4KN going up. how come? 

Permalink Reply by Rafael Pastrana on November 21, 2016 at 10:37pm

4 kN/m2 was rather an arbitrary value. The direction is Z, but local to the mesh. Have please a look with the Model View component.

Best,
Rafael

Permalink Reply by Alexia Ch on November 22, 2016 at 2:09am

I did have a look, but maybe it's me that can't understand the difference through mesh and model. If I applicate a load on the mesh is not as if I applicate it at the model?  :) my question was why upwards the load at the mesh? and not downwards? not z+ but z- so it can be a load of the snow lets say and not a pressure load that goes upwards. I don't know if I explain myself....