HW7. Interactive Geometry Optimization
Assignment
This problem set is asking to explore the tool developed by the Architecture department of MIT called Structure Fit, start using the parameters of exploration in order to, by choosing a first structure, explore the design space controlling the generation size and mutation to receive structures with a better perfrormance. Also, every structure cna be selected and taken as a reference or even you can select any structure and know its nodes coordinates. We are asked later to compare this tool with any other we can replicate in GH - goat and study and compare the diversity of those catalogs!
Something wich is curious to me is that for the fisrt time to me, there is a aesthetic criteria!. Im excited to compare how aesthetics affects to performance.
a_ Interactive evolutionary optimization with StructureFit
_i
We are asked to explore this iterative design generator and make three families! My three strategies are going to be:
Free Strategy
Free Strategy Conditions
Mixed Strategy
Mixed Strategy Conditions
Performance Strategy
Performance Strategy Conditions
As it can be seen, the most various design comes when I dont have any performance criteria. The solution came with frame structures with more variety of lengths, generating forms that are not optimal for load distribution, but very beautiful. On the other hand, when I have been only influenced by performance, the result are more repetitive design with more homogeneous tessellations that drives the load way more smoothly.
_ii
Plot
_iii
Catalog!
_iv
Computing divertity.
Based in the paper Mueller and Oschsendorf, it has been computed in Matlaab the catalog diversity obtaining a value of
Original Diversity = 516.75
In order to know wich are the designs that affect most to the diversity, I have calculated the diversity of the catalog by removing one design and analyzing each time how has change in comparation with the original. The result wich diferes the most with the original is the design that afect most to the diversity metric
As it can be seen, I underlined design 3 that is the one that affect most to the diversity.
Another way to measure diversity, It can be , instead using the max distance, we could use the minimun distance compared with the mean value of the distances.
_v
I have subtituted two designs to compare how it hsa affect to the total diversity
The first attempt has been subtituting the one of the two most boring design that affect the less to diversity. The candidate has been this
This replacement has increase my diversity to the value of 584.62
The second attempt has been to introduce another change, in this case the element subtituted has been the 10th design
The value of diversity now is 635.228
b. Now, in GH!
i
I have replicated a gh scrippt that allow you to move with sliders the 6DOF that the symetric bar structures has. This first point ask to introduce the original bar structure and optimimze it to chech how StructureFIT vs Grasshopper works.
This is the result of the optimized original. The performance of the optimized structure is now 0.63 with a stiffness to weight of 43.63
ii
I have tried to replicate some design that I made in my catalog and optimize them to see how the shape and the perfomance modifies.
The first candidate was design4
Original with perfirmance of 0.65
New with perfirmance of 0.616 using goat
The second candidate was design7
Original with perfirmance of 0.68
New with perfirmance of 0.59 using goat
iii
I have generated a brand new catalog that can be seen bellow. The value of diversity onve runned in the Matlab Script has devrease to the value of
New Catalog Diversity = 466.45
My workload has been generating a random structure I like, optimize it and then include it to the catalog. Maybe, this optimization makes my design more similar between them and this affect to my diversity. The diference with Structure fit is that in StructureFit I choose a form with a performance value, so my choice is more clean. Here, first I choose, then I optimize and the result is what I have got, making my choices more similar between them
