Good ideas and conversation. No ads, no tracking. Login or Take a Tour!
I've always wanted to see this kind of development in action. Awesome!
I've always found GAs insanely interesting. I love the fact that they can come up with solutions that we can't even comprehend, like in NASA's famous antenna experiment or the one where they evolved an FPGA that did tone recognition without a clock source (and this one, of course)
–
Thank you for the impetus to look up the NASA antenna experiment. It really is fascinating that functional design can now be taken to an entirely new level.
My recurring thought as I watched : is it possible to ensure a greater variety of shapes at the outset? They all seemed to be too 'full' to be likely blade shapes. For instance, it looked like an evolved Lenze blade was nowhere in sight. Widening the starting parameters helps to prevent 'lock-in' later on to less desirable shapes.
Most of all though, thanks for making this, as I've yet to find the time and determination to follow through with it ;)
–
99kiplingst · 4543 days ago · link ·
"My recurring thought as I watched : is it possible to ensure a greater variety of shapes at the outset?"
Do you mean they should be drastically varied from one another? Or just less less "full"? What would be the advantage? Would this just delay the inevitable or do you think there would be far more varied final shapes?
–
In particular the radiating spokes model seems too limited. If I'm not mistaken, the parameters here are any shape that would fit inside the circle, unless it needs to be attached in the middle. Even so, radiating shapes are just one 'family' of shapes that meet these parameters.
I'd have a go at programming the random shape generator, although I don't have any experience with wind simulations.
–
I wouldn't be surprised if there weren't a more-efficient non-radiating spoke model. However, I bet that would be a bear to program. The spoke/non-spoke models might be effectively orthogonal in terms of energy vs the type of motion, so you might end up with some sort of weird less-efficient hybrid.
What does a Lenze blade look like? I couldn't find it in Google images. Another limitation, is that this is just 2D. I imagine that a more-efficient 3D blade might have a very different profile.
thenewgreen · 4858 days ago · link ·
"Yes, shapes are prejudice... and they seem to do best mating with their own kind" -Does this eliminate the potential for an anomaly to create a genetic shift? -Very interesting work.