Creating Continuous Pillowed Surfaces on Pill / Racetrack / Bullnose

Hey everyone!

Looking to see what are people’s preferred ways to create the geometry shown in the screenshots; the pillowed surface in the end of a pill / racetrack / bullnose

I’ve found building half with NetworkSrf tends to pinch corners or not be tangent even when trying all the contingency options. Sweep is usually ugly most of the time as well

In the example below I split the arch curve in half [Red and Orange], and use the Surface Edge [Green]. I click in order of rainbow.

Rhino 7 file below:
pillow_bullnose.3dm (195.3 KB)

Could you re-rationalize this to a revolved shape that’s trimmed off, or a revolved shape that’s trimmed off with an edge surface?

This is a common enough question, and the fact is the way it’s done ‘for real’ by the company that is(was?) famous for such surfaces is brute-force point-pushing until it’s perfect, there’s no “tool” that has the logic required to just magically make such a shape without a flaw somewhere.

The bullnose end is not a perfect radius, so revolve isn’t suitable. Sweeps also don’t seem to a give result that’s as close as original post. One struggle is really why does NetworkSrf not work the way it should in this situation?

Using 3 Surface Edges to NetworkSrf. The preview looks nice and logical…

But it breaks when joined… surface is not continuous whether it’s on Position, Tangency, or Curvature

Feels like NetworkSrf is overcorrecting the tangency
Played with a lower tolerance and that doesn’t seem to make a drastic difference either

As I said there’s simply not a solution that lets you input some curves into a command and just get a perfect result from all angles without any post-point-tweaking, so something that makes THAT part easier would be the way to go. And as I said, the way Apple CAD sculptors do it with Alias is just what I said.

Hello- I usually like this configuration - the surfaces could be simplified but even so it is not bad -

-Pascal

2 Likes

Several approaches to similar geometry: Boat hulls with smoothly curve stems - how to model

1 Like

Can the edge curve be simplified with lower degree and fewer control points? It is currently degree 9 with 13 control points.

Can the arched surface be revised to be non-rational, for example degree 3 with 5 control points? It would then not be a mathematically perfect arc but very close.

The result of NetworkSrf is always degree 3 in both directions and non-rational. It usually cannot exactly match curves and edges of surfaces which are higher degree and/or non-rational.

Recommend to use one curve to describe the shape, and use Sweep 2 to make the surface,Then you could edit the control point of surface to adjust the shape.


pillow_bullnose_Ref.3dm (707.3 KB)

Hi, this is the technique I usually apply for such situations:




Note how well the CP’s flow. Everything is single-span.

Hey — didn’t see this response until now, thanks!

This seems like a good method for sketching out ideas. What is the purpose of copying the red surface twice? It seems I could just split the surface in half and use the two sides to BlendSrf?

And do you have a method of getting an exact dimension for the length with the BlendSrf command?

Hey Pascal, how are you extruding your green curves in the final poly?
I seem to get a break in tangency even when using NetworkSrf and having tangency to surfaces from the green curves

Sweep2 showed a break as well

It’s a standard technique used in class-a. If your hole is smaller than tolerance, you have a valid surface. Occasionally you see them in production models.

You delete the orange surfaces. You keep the red. This way, you don’t subdivide just for blending purposes. If you copy the red, you have a temporary backup.

No, because you create a singularity causing issues when blending. You should prevent singularities.

Not in Rhino, but you can estimate it… Anything below the tolerance is ok.