I’m building a Rhino model of an aircraft, a Sea Vixen fighter, as the basis for building a radio controlled model, and I’m having trouble generating a curvature continuous surface around the rear of the fuselage.
All my attempts end up with either an edge discontinuity, or some surface anomaly such as reverse curvature or a wrinkle.
Attached is the file with the construction lines and the surfaces that I’m happy with.
Is there a way to fill in the missing areas with a smooth surface?
Splitting it up into 4 smaller sections is the right idea, adding more curves to describe the shape you are trying to make wouldn’t hurt either. Also, if you have tried a Sweep2 the “add slash” feature would assist with this. If none of this works, I would rebuild the curves using Rebuild
I can blend this wrinkle out when I 3D print the model and finish it to use as a plug, but I’m trying to improve my Rhino skills in building a better 3D model.
It is a bit tricky to get that one surface to sit nicely on the trimming curve though.
@pbloxham Here’s the file, for what it is worth - I simplified some of the existing surface, quickly, I do not know if they are really done or if that was the best way to do it, but it will give you an idea.
Can I ask how you derived the shape of the untrimmed surface that gave the required trailing edge cut line - was it just trial and error, or is there some process you followed?
This is indeed nearly the right way to proceed, just not quite in the way you did here.
The problem is that you matched the auxilliary curve to be curvature continuous with the border curve of the adjacent surface. That causes the “wrinkle”.
But there are more fundamental problems.
As @theoutside mentioned your surfaces are way to cv heavy and you seem o want to create complex forms out of single surfaces instead of multiple matched surfaces.
You need to start from much leaner, cleaner curves.
Here is something I put together quickly:
(I simplified the main bodysurfaces and re-modeled just a part of the body. These new surfaces are not an exact match, I just wanted to show the principles)
This is far from perfect but shows you the way I would approach this problem.
The youtube tutorials @theoutside recommended?
Do yourself a favour and watch the whole series!
I’ve started to watch those videos, and it is helping. Another resource seems to be the Better Living Through CNC YouTube videos.
Some background, I’m starting with a 3D scan of a 1/48 scale plastic model which I then scaled up to 1/8 scale and then sectioned and used as guides to develop the surface guide curves. I’m trying to develop the guide curves to within 1mm or less (preferably less than 0.5mm) relative to the scan sections, in order to preserve the correct OML. To get that fidelity on a 1/2 section guide curve, I had to re-build them with 32 points, but that breaks the single span rule.
I have tried to split and simplify some of the guide curves in order to generate multiple single-span curves, but I haven’t achieved success with that strategy yet whilst still preserving the overall curve accuracy.
Further advice on the whole process would be appreciated.
Perhaps I’m being Captain Obvious here but consider:
Back in the 40’s and 50’s some designers at de Havilland crouched over their layout table and using splines and weights laid down the lines for the Sea Vixen. Using splines enforced a smooth continuously curved surface. Somehow somebody managed to scale those surfaces down to a 1/48 scale model and managed to maintain the same general smooth appearance. Then somebody ran a 3d scanner over that model which collected a bunch of points with the accompanying inaccuracies and noise inherent in the process. Then you multiplied all those inaccuracies by 6 as you enlarged to 1/8 scale. Then you sectioned to create wiggly guide curves which you now want to match with Rhino so you build curves with a whole bunch of points to make it go through your scan data.
This is where the fallacy occurs. Your guide curve approach should keep in mind the original designer’s methods and intention. Put your priority on using the Rhino methods which are analogous to the original designer’s tools, duplicating their original intentions rather than the scan data; use the scan data to “guesstimate” the original intent. As a friend of mine who made a model of a 1920’s Great Lakes cruise ship using only a high res photo for details said: “the ship doesn’t exist anymore and there’s nobody alive today that remembers what the details looked like, so I’m just going to do the very best I can”. In your case there may still be a Sea Vixen or two around and if accurate detail is important you should probably make an effort to visit one. You’ll probably be surprised at what the actual manufactured surface looks like, especially after 70 years.
Please forgive me if you’ve already come to these conclusions. Perhaps this will be useful to some others who are just starting this kind of project.
I have the Einscan SE scanner. I used the provided EInscan software to generate the scan mesh.
I was very impressed with the smoothness of the result and the fine detail that it resolved, such as the scribed panel lines which are only a few 0.1mm’s wide and deep.