NetworkSrf fails to maintain profile edge radius

in the attached file, I have found that using NetworkSrf on curves A,B,C,D sees it fail to create the radiused curve at their aft tips.
I tipped the shape (grey portion) using sweep2 and that went ok. Using MJ and MK as rails and point option picking point M.

What am I doing wrong to cause this to happen ?
I only get two of the four annotation dots (not mine) that appear in the help video. is that the cause, two columns are greyed out.

I am in fact trying to see what the curve at QRS would be, given the known aerofoil profiles A,B,C,D shown in the file.

Is there a way of networkSrf…ing the entire shape as one, rather than using a mix of NetworkSrf and sweep2rail ?
I wonder if I would get a different result when sectioning QRS.

I tried a few times but it failed. I tried with profiles split into two halves , upper and lower, also with them complete.

I can’t fathom out what curves I need for NetworkSrf for the entire shape, which should be closed, which open etc.

Surely networkSrf should be possible with a full curve I to H via KMJ…and E and F and profilesA.B.C.D or is it that it needs E to go all the way to R, ditto F. If so I am in catch22 land as QRS is unknown and the purpose of the exercise.

NetworkSrf trailing edge radius failure.3dm (278.4 KB)


Yeah… it does not, but in truth, this is not what NetowrkSrf is for, IMHO…this should be made with a top, a bottom and an edge (radius) surface. All three can be networks - that is fine (though with clean, matching curves simple lofts might be better).

  • NetworkSrf will want to be curvature continuous internally - these curves are tangent at the radius, so it will miss or wiggle due to that alone, most likely.

  • NetworkSrf refits all the input curves as G2 curves internally, - it does look like it could do better on these BUT

  • It really is better to avoid making a single surface be both a low curvature and a super high curvature surface all in one, especially if the transition is only tangent as here.

-your section curves could be better- they are not very consistent - they reverse curvature in (I am guessing) unintended places.

Wing_Steve1.3dm (44.0 KB)

You can also help matters and keep it a single network if you add some curves as in the image (black curves), rather than use the trailing edge curve - make sure you snap (interpCrv) to the ends of the arcs there.


Hi Pascal,
Understood on the weakness of NetworkSrf to use a tight and loose curve in one shot…

You are saying to treat the aft edge as a separate surface, sweep it or loft it after having done the top, the thing is I dont have a curve at QRS to sweep it to, as I am trying to establish with those I do have what curve would exist at QRS.

I wanted to create the entire shape in one go if possible to see what I got at QRS. so creating this in 5 or 6 pieces sort of defeats that task.

If I do top then aft edge as you say, and I also understand the two rails may be better than one, now that I have a top to make and aft edge to mate with it. then similarly do front, how do I get to do the tip ?

a separate mission for the tip with sweep2rail as I have done…is that still best approach….or can I do the entire shape another way.

I had wondered if I do the entire shape, and I am wanting to see what QRS would be, having no aerofoil profile for that section. would I see a different profile to how I have currently done it ?

It is not so much an exercise in filling spaces with surfaces, that is not the primary mission though they are needed ,…its about establishing what QRS is as a profile given the shapes I currently have.

I wanted to create the entire shape with one method, loft or sweep or NetworkSrf and see what I got at QRS.

I created E and F as before these two were made, when I did a sweep2rail, it looked like it was doped linen arcing down slightly between the profiles as on fabric aircraft !

After E and F were added in , it became a candidate for networkSrf in my mind, it looked so smooth and nice, bingo I thought.

You say I need to create a split on the profile up front as well and sweep that arc separately, so the upper half would be three separate sweeps ? There is no radiused curve there so do I just make a cut somewhere ? I guess so. Maybe in top view offset in from leading edge by 20mm or so.

I am unsure of what I should do for the grey area.

I opened the 3dm file straight into rhino rather than save it then open it as i normally do, but choosing shaded I see no surfaces. I dont see a finished shaded shape, just some lines.

The curves are from tabled data 70 yrs old so may be a bit off. but if I start altering them then I am altering the result of the test. I did alter one using curvature graph but it then impacts on another which shares its rear half so I had to alter that.


Sorry- no surfaces- here’s how I’d divvy up the curves - you can join the section curves into single loops and make a single surface as well- that will work.

Wing_Steve1.3dm (52.0 KB)


I see same file as first but with the twin rails at rear, …I correct my writing, you say divide the curves ABCD up into two, a separate rear portion and the remainder as one, as per your attached file , …somehow I misconstrued and thought also the front edge needed cutting out and dealing with separately.

See my reply which I was typing when you were doing this,

ok, just tried that, however InterPCrv result does not follow curvatire of original aft edge line in a parallel way, so the resulting surface would be less than the correct shape. The radiused curves were not a standard radius so just offsetting the curve (rail) inwards is not an option.

Should I sweep them along the single rail, then make rails from their surface edges, then use those rails to re-sweep them, well they are made by then ! Use those rails instead to form back edge of the upper skins ?

its all about establishing QRS shape as a profile curve, additional to getting this all skinned up.

skinning the whole thing in one hit using these existing known curves was my idea to see what happened at QRS, then I would section it to get profile QRS.

At the mo we have all but that area being skinned better.