generative modeling for Rhino
When i first set multiple curves in rhino curves (5 to 10 seconds in the video you gently rec) i don´t get the surface (the curves become pink but the lofted surface is not there).
maybe this initial error when i start "IO generated 20 messages, would you like to see them" is involved?? True is i´m very interested in test your work. Hope I could test it finally... :P
Second: Agree with the 2.34e-10, but you know, corten steel is expensive and the surface cover is large (1000m2), i would want to know "before buy" if that dev surface is possible with the NO stretch caracteristic of steel. Maybe i can assume that difference (welding might help)
*BTW, how do you know is exactly 0.06% bigger? Sorry for the noob question
Glad you could work it out. You don't really need it since there are only doublons in the case of cylinder, when a generatrice joins the ends of the directrices.
about the speed, I guess it depends on your machine. the population size depends mostly of the number of genes. I do it with 30, but maybe less could do.
Note that Sampling is a very sensitive parameter, especially with weird cr***y curves like synergy's ones.
You can't really expect to find a nice fair developable surface between to random curve. They need love and passion. You can acheive quasi developable surface with low sampling and cusp filter on. but it's not right.
This galapagos set up only modifies the weights of the control points, but you're supposed to work a little on cp themselves before or/and after using it. Galapagos will then "fair" the planks, balancing generatrices and hence maximizing Min Radius. Usually 5 to 10 generations give good result.
Actually, to me, this galapagos thing is more fun that useful. I prefer to control manually the curves, keep track of tangencies, curvatures,bow and aft silhouettes,etc. Anyway, I feel like galapagos is a dead end here. I'll try kangaroo, or perhaps Daniel will be interested....
well I don't know about corten, usually thin metal sheet locally accept some double curvature, and even thin plywood to an extent. It depends on the material plasticity, plank thickness, fabrication techniques an tolerances. Up to you really...
It's quite easy to build physical models, then you can "feel" the tensions in the sheet.
This is why I'm interested in kangaroo. It's a bending resistance issue after all.
Or you can split the problem an do someting like that.
About these IO errors, it's surprising. did you install karamba? it's required for now, or you can fix inside the cluster (see previous posts). do you use gh latest version?
I'll try to make another version for next weekend.
the % bigger is an output of the UnrollSrf command in rhino (in the console).
Sorry, Ill be late for the new release. I found a lot of bugs I wasn't aware of, plus all those I knew of
Not fully functional yet, but I added a relaxation kangaroo tool for auto-fairing curves. Funny.
Checkout the wip teaser.
Gret work Fred. Thank you so much for your effort.
I think a developable surface solver with a referenced surface (lofted lines)
will be a VERY usefull for us, architects.
Can we test it?
I tried the kangaroo option with 0.00 deviation (think is an easy way to achieve near zero gaussian curvature isn´t it? a better way? with a curvature range of 5.28 e-10.( are this in mm?, unit? I´m drawing in meters) (0.44% bigger).
Advice reducing the gaussian curvature?
*the pc become slower.
The kangaroo tool does not guarantee a developable surface. It's useful when you don't know what the surface could look like. When you have no experience with dev surf, it is in fact difficult to know how to modify your curves.
But if the curves are too weird or too different, it won't get much better. Remember you can still move control points while kangaroo is running.
The wrong of this method is that you don't have real control.
Setting deviation to 0 is usually not very useful unless your curves are almost good. With the wavy curves you sent me, I was happy with 2 (Deviation is something like a % of the curve length).
I much prefer the other method. Curves with few control points, you're able to define tangencies, you've got high degree continuity. With 4 to 5 cp, Galapagos will have a wide range of shapes to explore. With more cp you'll get smooth tuning. You can still take the relaxed surface as a base to draw such curves.
Other ways to increase developability.
First is obviously Sampling. This one is quite tricky. Higher sampling might result in either better or worse results. It's very cpu expensive too, because it involves a little more than (Sampling)² operations each time.
Second is the surface type. Usually Network surface is the best, and allows to join multiple planks with less sampling. It's more expensive too, and runs only with Rhino 5. Loft tight is a decent replacement.
The "Reverse curve list" button, and the multi planks mode when you input more than 2 curves.
But everything depends on your tolerances and that I can't answer.
There are still some bugs here and there, especially in the cusp filter part. Sorry about the unit of Gaussian curvature, I forgot to fix it. It's model units, so it should be [1/m²] in your case.
About the pc slowing, it's really vital to block timers all the time and turn phasma on and off when you must. I experienced some slowing today though, and I was worried, but it's normal now so it must have been Windows or some GUI thing...
Any success anyone? I'd love to see a few developable things here... Give me some love, come on!
Working with it...
I love you Fred!!!
I´m already testing it, i give you more feedback soon.
thanks a lot for your work Fred.