Galapagos Convergence

Hi all,

I am trying to find the optimum shading depth for a profile that can be 1,2,3 or 4 feet deep outside and inside using Galapagos (4x4 = 16 options total). I have defined the fitness function as 'Maximum of (Well Lit - Over Lit) illuminance nodes). The results are not converging. Any suggestion towards a solution will be much appreciated!

Shamim.

Attachments:

SJ_mov_03nov.mp4, 281 KB
SJ_ScreenShots_03nov.pdf, 588 KB

Replies to This Discussion

Permalink Reply by David Rutten on November 3, 2013 at 10:38am

Hi Shamin,

there are so few possible different genomes that Galapagos finds the best one in generation 2. After that it keeps trying to find different ones but it can't, so the solution doesn't converge any further. However if you stop the solver it will give you the best answer it came up with.

David Rutten

david@mcneel.com

Seattle, WA

Permalink Reply by Shamim Javed on November 3, 2013 at 1:01pm

Hi David,

Thanks for the comments. I have increased the complexity. Here is a light shelf I am trying to optimize for 'good daylight' [fitness function: maximum of (well lit - over lit) number of nodes]. I have 5-variables with 4 options each (a total of 4x4x4x4x4 = 1024 variations):

Height: 6,7,8 & 9 ft

Depth inside: 1, 2, 3 & 4 ft

Depth outside: 1, 2, 3 & 4 ft

Rotation outside: 0, 15, 30 & 45 degree

Rotation inside: 0, 15, 30 & 45 degree

I do not see convergence yet (see attached). The last combination of variables say 9-ft high, 4-ft deep (inside & outside) and 0-deg rotation (horizontal) inside and outside. Is this the optimized value of the light shelf parameters for the best illumination level at the sensor plane for that specific place, time and date?

Shamim.

Attachments:

Permalink Reply by David Rutten on November 3, 2013 at 2:02pm

I see. I think maybe there's a confusion about how Galapagos works here. Galapagos doesn't care about converging all the genomes, the only thing it tries to do is find a genome with high fitness. As long as it keeps finding better answers it will continue to run, whether or not the rest of the 'population' is converging. Similarly, if it cannot find a better answer in -say- 20 generations then the solver will terminate, whether or not the rest of the population is converging.

The only thing that matter are those little [+] symbols that are drawn when the yellow band goes up.

David Rutten

david@mcneel.com

Seattle, WA

Permalink Reply by Shamim Javed on November 4, 2013 at 5:57am

Thanks David for pointing out these basic concepts as I try to learn this tool. Do you see the potential of Galapagos to help me do what I am trying here, i.e. optimizing a light shelf for certain parameters (in this case 1024 possibilities) for a given time and place?

As a next step I would like to find an optimized solution for a season - say one position each for summer, winter, fall and spring. Going further, I would like to investigate how this optimization is modified if heat gain/loss is brought into play. The incorporation of natural ventilation considerations can be an additional layer of variables for the future.

If anyone else in this forum has done anything similar I would appreciate some guidance from them as well.

Shamim.

RSS