Increase optimization speed

Cobertura metá (1.0 MB)

How can I reduce my file so that the optimization with galápagos is done faster or what can I do so that the optimization process is not too slow? …
I hope you can help me; I upload my file to be reviewed …
Thank you

Do you know what Galapagos does in background?

It basically checks random values on all sliders. Genetic Algorithm. The sooner it converges to a value the shorter the run. Result accuracy is not guaranteed.

Ways to improve the runtime:

  • Use smaller ranges in the sliders
  • Use less sliders
  • Make the slider value co-dependable. Understand, similar to the upper point. remove one slider and substitude it with a value taken from some other slider with formula.

Hi there, I wasn’t able to fully understand what your definition does as I don’t have Karamba on my machine. However, it looks like your definition is doing a lot at the same time. I think a general shape is being described but some detail members are also worked out at the same time. A picture of the geometry would help.

Your problem is that the scope of search is too big and the calculation time for each attempt is probably very long.

@ivelin.peychev had some good recommendations about the sliders. Remember that for each slider you’re adding, you’re increasing the dimension of the solution space. You go from looking for a point on a meter stick, to a point on a map to a needle in a haystack, etc. I also found that if the algorithms that are affected by sliders are smooth in their change, the solvers have a much better chance to converge on a solution.

I would recommend breaking the optimization problem up into bite size pieces. I’d start with a very simplified structural setup with only the key members in place. That definition should be able to run really fast and be controlled by only a few variables. From there you’ll have a good idea where your final solution will lies so you can either lock down those parameters or significantly limit their range. From there you can run subsequent optimizations where you increase the complexity. Keep in mind the the maximum complexity that can be optimized in a reasonable amount of time might force you to keep trivial parts of your design out of the optimization process and leave them to be strictly dependent on other more important aspects.

My take on your definition is that you are trying to assign different sizes members of trusses to a set curved frame. If I am correct, you could break down the structure into parts and do them directly. Without knowing your design completely could you design the members spanning between the frames before you design the frame trusses?

other option would be to use the karamba optimise component to get a close guess and then run your algorithm. I think the main reason you don’t is because of EC3 Vs AISC, but for a preliminary this wouldn’t matter, you could even back off the karamba ptermise sot hat you have a 80% section ect. .

Hi all.
I’ve already reduced the sliders and ranges, but it’s just as slow. I think the optimization will be done in parts.

What I want to do is find the geometry, the section and the number of armors whose weight is minimal, but each element that makes up the armors meet the design requirements of the AISC standard for tubular sections, that is why some scripts to design with that standard. I wanted to optimize everything together, but my optimization function is very long.

Thanks for your comments, I will try to do it by parts.