algorithmic modeling for Rhino
I'm working on a project to optimize an arch structure for multiple load combinations. I use Kangaroo for the form finding part and Karamba for the structural analysis.
Karamba analyzes the structure and gives a normative bending moment out of one of the load combinations. Galapagos is used to minimize the maximum bending moment in the structure by varying the horizontal load vectors in Kangaroo.
Galapagos finds an solution at the end, however, the most optimal solution is not always the last iteration. When I select the most ideal value and click on reinstate, the structure doesn't go to that value, in fact it changes the values again.
How can I solve this problem? Where am I going wrong? I practiced with simple models and they worked fine. Then structure changed to the value, after using the reinstate button.
Tips and help are very welcome!
(If you want to try the model in de file: Before starting Galapagos, first start Kangaroo)
I think it is because the way you use Kangaroo in the solution. When Kangaroo runs it often needs many iterations before it rests in a stable position. Galapagos does not give time to Kangaroo to find that stable position before it changes the values of the sliders.
When you stop Galapagos, Kangaroo will have time to run some more iterations and stabilise the structure. Since the structure changes the results of the analysis in Karamba will also change. Notice that the results in the gene list will change when you try to reinstate them. Click reinstate and then click on the gene list and the gene value will change even after Galapagos as finished the optimisation.
There may be two solutions. One is to use the ZombieKangaroo. This solver will run as many iterations as you want without being interrupted by Galapagos. Another solution is to first make a shape with Kangaroo, and then let Galapagos and Karamba work on the "frozen" shape without more interference from Kangaroo, but that's maybe not what you want.
Thanks a lot for your fast response. I tried the ZombieKangaroo component and the 'reinstate problem' is gone. You were totally right.
Another problem I'm facing is the final solution Galapagos generates when it tries to minimize the bending moment. I ran the optimization several times and every time, the solution is a bit different (figures below). Is this normal if you have relative many variables? I have 11 sliders which Galapagos can vary.
I start everytime with the same settings and start points.
Or is it possible to get the same answer everytime? That would be nice for my model.