Corner Fillet Construction

I’ve been asked to construct a corner fillet in the attached. What makes this one a little different is that the solution has to retain as large a portion of the 38mm radius sphere as possible in the corner fillet surface. Edges marked 1 are arbitrary and can be adjusted as required. The edge of the small fillet surface (2) has to stay where it is. The two 38mm fillets aren’t set, but I’ve aligned them tangential to the sphere. The missing surface has to remain convex throughout and must finish tangent to any retained portion of the sphere and other boundary surfaces.

I’ve tried a few ways of generating a solution, but so far it’s defeated me. Anyone here got an idea?
Corner Fillet.3dm (315.6 KB)

I took a look at the file and I’m a little confused. What is this detail and why was it modeled this way?

Hi Matt, I don’t quite love how this looks yet, but it might be an approach, if you have not tried it yet. The white curve is on the sphere, and the green one is a blendCrv - you can fiddle with the controls and slide the end along the white curve until it looks nice, to give you an idea where to chop the sphere off and get a reasonable progression in the surface. I used an InterpCrv (hidden) to trim off the sphere, then made a ‘reminder’ curve (cyan) and networkSrf…

Corner Fillet_Maybe.3dm (106.0 KB)


What’s shown are in effect the inside faces of an aluminium pressing that forms the corner piece of an underskirt on a train. I’ve got to recreate the part from the original paper drawings, drawn long before CAD existed. The drawings are very vague, citing pretty much just what I’ve laid out. The transition form that I’m struggling with appears only in one view, where they’ve simply drawn in the 4mm and 38mm rads and given no further information, other than a comment about the corner fillet being formed as part of a sphere of 38mm radius.

I’ve seen one of these ‘in the flesh’ and the corner does indeed look like part of a sphere that then blends back into the tight corner rad.

If I could figure out a way of making a blend surface that starts at the tight radius and ends as a tangent to the sphere, I could do it - but I can’t! {:o(

Thanks Pascal. An interesting approach. How did you create the white line? I was doing something similar with the sphere, revolving it until I got an isoparm pointing in something like the right direction, then extracting it. That proved pretty tedious though. Was your ‘reminder’ curve constructed manually?

It’s still looks a little ‘pinched’ at the ‘1’ boundaries which is a problem I was getting. This could perhaps be eased by backing the ‘1’ edges off a bit (which I’m at liberty to do), giving the network surface more room to change direction smoothly.

I’m not quite sure what it is about this chestnut that’s difficult to crack; when I first looked at the original drawing, I dismissed it as being trivial!

I used ShortPath.

Yes- that is a bit tedious, to be sure. The end of the extra Sweep 2 that I added off the end of the exiting fillet could be better - more arc like - as a base for the network srf there too. This one is maybe a little better.
Corner Fillet_Maybe2.3dm (110.9 KB)


1 Like

I filled in the corner with blendsrf. The only trick to getting a good blend is to figure out where to make the cut in the sphere.

CornerFilletX.3dm (62.7 KB)

I just looked at Pascal’s solution which looks similar

1 Like

ShortPath. That’s my new tool for the day. Didn’t know it existed - thanks Pascal!

you have to get ahold of the paper or jig etc used by the dude actually making it instead of the original drawing. :wink:

how were they made? do you know?