Select two curves and a rail and it just doesn't work right. I've rebuilt the curves (not the rail, should I?) without success. And FWIW, I could not get the variable radius to work on this edge either. Thanks for any help!

FILE: HERE

1. Second part of sweep isn't tangent to surface, but both originating arc curves are tangent.
2. Sweep settings.
3. Another view of the incorrect sweep.

Hi Chipp, when you sweep on a rail that twists around in 3D space like your rail here, there is also a banking effect that happens which means the profiles will tilt left or right around the curve tangent as they step along the rail.

That banking combined with the additional rotation of taking a slanted rail that then tries to keep the same slanted position in relation to the rail as it steps along it results in a lot of sort of frenzied motion of your shape near the end.

One way to reduce that is to use the "Twist: Flat" option in the Sweep command options in the last stage of Sweep. That will only rotate the profiles around one single axis direction (by default the z axis direction, which works ok for your case here, otherwise use the "Set flat direction" button to specify the allowed rotation axis). If you use that it will sort of calm down the amount of rotation that is happening between each step along the rail and you'll get a better result than your original attempt.

But really probably one rail sweep is not going to give you the kind of result you need here because it does not fix down the side edges of the generated surface to hug right along the side surfaces.

You might try forming some side rails by projecting curves onto the side walls and then use 2 rail sweep...

- Michael

The reason fillet is having trouble on your shape is due to the sharp sort of "upsweep" shape right near the end:




That fairly sharp bend in shape near the end causes the fillet to need to mutate a lot in a short distance, things like that are generally not so good for getting good fillets.

One thing that will be particularly difficult is to get the fillet to extend properly in such a way that it intersects cleanly with the back face here:



With the fillet swooping close to being tangent to that back face it becomes difficult for it to intersect well with it.

It would probably work a lot better if the shape did not swoop so steeply in such a small area near the end there.

- Michael

The "project and get side rails" way would be like the attached.

You can now select the arcs and run sweep and use the 2 open edges (which are projected trims from the 2 lines) as the rails.

The resulting sweep will hug the surfaces and will have the shape you want at each end of it, but it won't be guaranteed to be smooth on the inside areas. Generally only fillet or blend will yield something that is a smooth connection between 2 existing surfaces.

- Michael

Thanks for explaining. Very much appreciated. I'll take a look.

I ended up using loft to create the basic shapes. Then I tweaked the originating curves to the correct side/top profiles which was great as the lofted shape adjusted instantly as I tweaked. Super for 'fine-tuning' a shape. Lastly, I deleted the loft shape, added fillets as I liked and re-lofted. I tried adding fillets with the 'live' loft but things got weird fast. Seems like the "Samardac Method" is a great one for parametric-like modeling. :-)