Grasshopper

algorithmic modeling for Rhino

Hi everybody !

Sorry for my english, it's not my native language.

First above all, i want to precise that before posting this, I've read - almost - all the posts concerning multi-fitness solutions with Galapagos and that I'm already accustomed with mass addition of the fitness. But this can't solve my current problem. I'll tried to explain the best I can.

Goal :

I have a tower façade to cover with modules. Each floor is divide with a radiant structural grid. Let say for the exemple that each floor is divide in 9 parts.

I want to put modules on each parts in order to cover the whole length.

The thing is that each floor is different (so finding the solution for one floor is not usefull). So we decide to use three kinds of module :

  • Module A is 793mm long.
  • Module B is 893mm long
  • Special module as a size between 0.714 and 0.872.

Each part of the façade must be completed with 0 to 10 Module A, 0 to 10 Module B and 2 Special Modules. Special module size can change for each part.

Solution for one part

I've realized a Galapagos solution which change these tree parameters (number of Module A, number of Module B and size of Special module) in order to minimize the difference between the total length of the modules and the length of the façade's part.

This is working very well.

Solution in Grasshopper :

Before runing Galapagos :

After runing Galapagos :

The problem

I've now to generalize the idea for the total number of parts. In my exemple now, I'll show 9 parts but the real number is several hundreds.

I changed the Number Sliders in Gene Pools. As for instance, the first Gene Pool contains 9 integers for the 9 numbers of Module A (one number for each part).

I used Mass Addition for the fitness.

And then you can guess the problem : basicaly, the solution is working. But it's very slow. I need so many time fort 9 parts that I can't imagine the time for the whole tower.

Why this ? Simply because Galapagos doesn't understand that the number of Modules A for the Part X has no influence on the Part Y. So it tries, for each part, to change every paramater (3x9 in my exemple) unless to change only the three paramaters wich affect each part.

Thus, with a large number (600-700) of parts, it's impossible to reach the beginning of a solution.

Is there someone here to help me ? Please ? :)

Thanks in anticipate,

Marc

Architect and structural engineer

http://parametriclab.eklablog.com/

Views: 1753

Replies to This Discussion

always best to upload a file...

Marc it's very difficult to help you without the definition and rhino files

Thanks Omar and Ethan for the answer.

The thing is, I've not the autorisation to send the model of the project.

But I've done a file with only an exemple with everything necessary to the problem. You can find these files attached to this post.

Thnaks again in anticipate

Sorry, here are the files

Attachments:

Little up : nobody has an idea ?

Why do you not run it separately for every segment?

That would work and take less time then run the optimization with hundreds of Gens.

Or did I misunderstood something?

I know that’s not flexible, but I try to understand your problem.

Hi Bene,

Thanks for the answer.

The problem is that the whole tower is composed with 600 different segments. So the manual solution is not possible, unfortunately.

Cheers,

Marc

Yes, I can imagine.

One solution could be, that somebody write a script witch works with DataTrees and solves your problem.

Writing a simple Evolutionary Solver is not super complicated.

But I guess there is a more efficient way then evolution.

 

In conclusion, I don´t think there is a workaround with Galapagos.

 

Cheers

Bene

RSS

About

Translate

Search

Videos

  • Add Videos
  • View All

© 2024   Created by Scott Davidson.   Powered by

Badges  |  Report an Issue  |  Terms of Service