A few Qs:
The lines you created for curve blending are all obvious except for this one- Did you project a line across the object?
I’m not sure how you created the cutting surfaces in this instance. Was that a edgecuves untrimmed, or perhaps a sweep? The curves I think you used are joined.
Fig 1. For the upper U formed “BlendCrv guide” I used a leaning Surface (going right through the object) and called Curve \ Curve From Objects \ Intersections to “extract” this curve.
Fig 2. The short guides at the bottom I made with Extract Isocurve (the orthogonal ones). For the other ones I used Line \ Perpendicular From Curve (from the brown main Intersect curve).
It’s important with the angles, although one might not think´so at first… (but you have seen the mess you’ll end up with if not carefully studying where the “imagined curvatures” will tend to go.
Patched i think, That’s why they look like that (untrimmed- effect?, I don’t know).Edit: Oh, at the arrow in Fig 2., that was an old guide for a BlendCrv which didn’t do any good, so I moved it outwards a bit (the one pictured).
// Rolf
I’ve been studying your work. I think I’ve managed to understand it and I’ve an example. You can see the blend curves created and the blends made. Network surface creates a reasonably okay surface, but as you can see there’s a creasing. For some reason I’m not able to adjust the surface with control pts on the networked curves - that is to say history is on but not responding.
I’m also dismayed how clean surface generate so many control pts when used to create intersection curves.
And there’s still the issue with merging at the rear.
COntinuing:
I add two more control curves and one problem is solved and another created.
all told, it is way too much work to get something as fundamental as a smooth transition.
and then another attempt with carefully tangential blend curves
I suggest you extract an isocurve at the “middle” height of the crease (at the arrow tip in the last picture) and run FitCrv 0.1 on it, and point edit (with History) if necessary.
Perhaps 2, or even three such horizontal control curves are needed.
I have studied this some more, and the reason for trouble is, as always, the starting edges. When focusing on that, and strictly dividing (first in the imagination) radiuses and planes, and fixing the "start angles (as close as possible to 90 degree in the corners (as we have seen in other geometries), then the surfaces tend to be near perfect at first shot.
Also, important in this case; the curvature is so complex so that straight constructions curves (straight compared to the isocurves) does not tend give good results when the curvature is so extreme as here. For this reason, extract an isocurve and modify that curve instead, since they follow the isocurves better. You will immediately notice better results.
Edit:
The number of construction lines should follow the following basic rule: Start of radius, end of radius, next start of radius, next end of radius, and so on. Preferably no extra lines in between. Rather add crossing curves to “lift” or “press down” a surface that tends to go bananas.
// Rolf
Use FitCrv (0.1 mm) on the intersection curves. That usually gives the desired minimum number of control points. Finally, manually delete excess CPs, if any (preferably while in History, so you can see what happens to the surface, and so you can undo if the surface goes bananas)
// Rolf
I regret that you ended up with creases. It looks like you were onto something, and then those tensions screwing up the whole thing.
I have tried some different things, like starting with “strict structure” in the small and big fillets, and the try to fill in between. The intent as that one could just go “by the book” following a strict strategy, and although time consuming it still wouldn’t be very slow if one could just follow a set of rules and directly get a predictable result. Well, I thought I was onto something, and it almost succeeded, but then again I had to start drawing construction lines to support the surface, en there it was, endlessly time consuming again.
One problem is that the sides of the upper “bar” (what is it called?) are leaning a bit, and so the upper edge of the lower fillet (or the extension of it) tends to force the surfaces at the sides to curl or crease, especially if using Curvature continuity. Also the arc, or the “half crecents” on the under side makes the corner complex and which in turn makes the surface confused (it bends the surface along the extended fillet inwards, resulting in that crease you have, as well as wrinkles near the corners, aso). But at least we have become quite skilled in freeform modelling… 
(I spent days on this).
Anyhow, here some pictures showing what I tried. Must think about it more though in order to find a “straight forward” method that doesn’t require to support the surface (other than guide the edges).
Fig 1. Here I placed (3) Points (using Osnap “center” on the fillet) to make four short Lines (Line Through Points). The 4 short guide lines are selected in the picture. And then I BlendCrv:ed between them to get the curved path (covered with a pipe). In reality I used Sweep1 and the fillet profile as a cross section. Same thing on the inner side with that very big lower “fillet” ending up in the small upper fillet towards the hub center. The edges of the sweeps could then guide the surfaces between them (not shown here).
Fig 2. The final Sweep1ed fillet profile selected. As you can see, the attempt was to make very strict details, strictly starting from points of “drastic changes” in the base geometry, and just tap, tap, tap the mouse until finished. But … at least I advanced my skills a little bit when trying this. But in a year or two I might be there…
Fig 3. At the end of the day I had to support the surface with construction lines anyway. But at least I managed to avoid those terrible creases… (but you seem to have smoother surfaces in general):
So dear James, what to say?
I spent days on this and I must now continue with my own projects. For now I just wanted to let you know that you’re not alone out there. This one was really nasty, but eventually we’ll fix this one as well. I have a bunch of even trickier geometries to draw in the next couple of weeks. (iron cast chassis details for an old car).
Don’t give up.
// Rolf
Tsplines for Mac! Fusion 360 (free for a year) is available with mac through browser. Autodesk bought tsplines and it is now in Fusion 360. If I had to start fresh, I would use Fusion 360 for tsplines work and then import iges file into rhino for the stuff best done in nurbs-rhino.
Rolf, you’re a mensch! Thanks for such detailed help. I hope you’ll let me see the coming work you’re doing on the auto.
yeah, I’ve been looking the application over. I can see it’s value, but having spent the last year learning rhino I’m somewhat irritated by the prospect of another year learning another interface.
It is however a very impressive suite of applications. I think it’s probably going to dominate the product design industry in a very short time.
I think Rolf’s results are better than mine but I at least wanted to share my 2 cents worth after spending a little time on this today:
All of the above I learned in Kyle’s level 2 class, which is well worth the time and effort if you want to perfect these techniques.
Maybe a patch is better than the 2 rail sweep?
I can see your point, but you’re keeping the surface much shorter than is required along the spoke and that’s where the creasing problem is happening.
Hi,
I post a solution at XNurbs releases a ground-breaking NURBS software. For XNurbs, the operation is very simple, just select the curves. (For this model, to get the best result, interactive designing is needed. We are not designers, so we only use the simplest way to do it on the fly.) Is this what you want?
