Hello I would like to merge multple surfaces into 1 accurately. If i use smooth= yes it will change the shape, when I use smooth=no, it will keep the shape unchanged but if i merge multiple surface it will have problem, so last step I use rebuild, but it doesn’t follow my guide curve anymore. Do anyone have better methods? thanks
Mirror the halves so your sections are closed curves. Then loft
In some rare occasions _mergeSrf helps to overcome limitations of PolySurfaces.
I see your intention: To have more control over a surface define smaller sections and then merge them - in my opinion this is will cause more problems then benifits.
Either keep the initial patch-layout, or build a single surface as Martin suggested - also check the option “loose” of the loft command.
another option would be to build 4 surfaces / 2 symmetrical - the approach of having a “helper”-surface perpendicular to the later mirror plane is quite common. (as you did) (but keeping the patch layout, no merging)
instead of modifying the CVs of multiple curves you might benefit from building a clean surface and modify the surface’s CVs.
use
_emap
_zebra
_curvatureAnalysis
_curvatureGraph
to evaluate the resulting surface
…
_gcon for Curves…
also use curvatureGraph to check the curves:
hope this helps - kind regards -tom
A good suggestion - if - the exact shape between the sections is not important.
@VinPo I see that your intial surfaces are all single span of various degrees. But the final desired result is a single surface which needs to be multispan.
What continuity do you need? Curvature continuity? If so then curvature is not continuous between your input curves. That means surfaces created using those curves will have curvature discontinuities between them.
Yes this is very close to the result I want, but the loft surface doesn’t follow the rail.
What I use mostly was the network surface after patch layout is completed. Recently I am impressed by thirtysixverts with idea of single span clean surface “sculpt and build”. The task here is infact the fuselage around cockpit of F-14, there are lots of add on features on this fuselage such as panel lines/ engraved lines, vents, pitot tubes, rivet heads etc… all these massive amount of solid is going to trim that “primary surface/ base surface “. If that surface is not one piece, I will need to do lots of untrim/ trim works when client request corrections.Thats why I try to combine those single span surface into one. If it doesn’t work, I will just use network surface instead.
Another thought : maybe I can simplify the patch layout for one single span surface…
That yellow line you highlighted is infact the quad of cross sections and formed this line- this will be tangent of surface created- yes I use mostly tangent only, I think curvature creates flat area.
no.
if you just need a single surface for simpler workflows as you mentioned. try _mergeSrf with the smooth option turned off. (this will create a MulitKnot / “Kink” (which will keep the continuity before the merge - whatever it is/was))
I overlooked this end.
perhaps this change is acceptable per design intent:
this too:
one more:
Therefore, this is the network I think you’re after:
First rule of quality surfacing: never use network surface.
Or, nonsense 
.
lol maybe the iso’s made em dizzy.
@VinPo there’s no reason to be afraid of ‘ntwrksrf’s’ 
Imma put some time into the netwrk and finalize a draft.
I noticed these crvs are degree 7, and only really need to be degree 3 or 5 if you’re one of those deg 5 enthusiasts.
But since the other crvs are deg 3 and netwrksrf likes to only use deg 3, I’ll probably rebuild them to deg 3 prior to ntwrk.
After rebuilding the crvs, I netwrked it then rebuilt the resulting srf. Everthing is within a few tenths of a thousandth of deviation. Finally, I took a shortcut and mergedsrf into 1 srf.
Hence:
Some ppl might be scared of the iso’s.
Here’s a version accurate to a few tenths of a hundredth:
Here’s a subd version based on the initial dense nurbs:
Question 20240107_emod.3dm (18.4 MB)
Obviously the iso’s can be controlled differently if the design intends. And depending on the netwrk, it can always be reworked to the original geometry etc. It all depends on the goal of 1 single surface constraint, etc. The wiggly iso’s probably mostly came from the subd messin’ aroun’ which could be revised and improved as well.
Also the original netwrk could be analyzed, and improved from the beginning. Depends on the origin of those curves, and the constraints there might be to the origin.
If the desire is to simplify things then the degree 5,7 and/or loft gurus probably want to see very sparse iso’s under the belief that they’re easier to work with, manipulate etc, and thereby also deviate from origin etc.
The rule of ‘single srf’ has it’s pros n’ cons, like anything else. Generally it’s wise to isolate high curvature areas into separate srfs, but it’s always fun making them merged into single srf nets 
Surface quality (as I intended it) is not manifesting itself in minimal deviation from the profile curves.
If that is your primary goal, by all means use network surface.
However, to achieve a smooth surface with pleasant highlights a deviation from those curves is needed.
The chance that a smooth, high quality surface will go exactly (with zero-point-zero-whatever deviation) through the curves is effectively zero.
Again, it all depends on what is meant by “high quality surfacing”, so my bad for not defining this clearly.
Apart from that, I stand by my words: (one of the) first rule(s) to achieve smooth, pleasant surfaces, without unwanted bumps or inflections or discontinuities within the surfaces: don’t use network surfaces.
Edit/PS:
Of course, the thread title makes it more or less clear that your interpretation of surface quality is the one pertinent to @VinPo
That is what I always do. However this network surface is just like thritysixverts said, it is done by tolerance and when curvature graph on, there is chance to have some regions change from convex to concave, so I am looking for more clean surface (less point count).
Do you think if reconstructing the guide curves, such as deleting some cross sections and rebuild curves to one pieces will simplify the surface as well as keeping the same shape?
I always do curve network, but the surface has many point count and difficult to do shape edit. Primary surface is what I learnt from thirtysixverts: sculpt and match edge curve surfaces, but its 10 pieces of clean surface, merging them with smooth= no will have multi knot on the surface. So when compare these 2 options I will still go for curve network.
Well the best way to reduce the surface density, is to isolate different curvature inflections and avoid making 1 single surface. But it depends what the objective is, cause there’s obvious pros and cons either way.
The network curves are the base starting point, and will lead to potentially high control point density depending how closely the netwrksrf conforms to them. While the resulting surface can be edited and rebuilt to a lower density afterwards, etc. I suppose one could explore rebuilding a higher density deg 3 UV to a lower density deg 5 UV…
The drawback for netwrksrf is it’s only seemingly able to create degree 3. If in the future it were able to be degree 5 compatible, then maybe it would be more compatible with lower density results with better “quality” upfront – as some might say.
I personally don’t see any reason to use any degree other than 1, 2, or 3, on any given surface.
Although, I’ve seen some users demonstrating some interesting things with deg 5’s.
If networksrf becomes compatible with deg 5’s someday, then I’ll probably play around with it more.
My opinion is, it does take skill and patience to successfully manage curve networks that are necessary to create elaborate, organic, high compound curvature, high density surfaces. And that is what leads to particular users being against using it. It’s not easy when you have a situation where there’s maybe 50 or 100 or 200 curves and something isn’t right and the surface wont network right away and you have to find the problem – like a needle in a haystack per say.
But, believe it or not this workflow does have a skill curve, and the newbs are usually the ones that will hate on it and act like it’s not them and they’re being the newb isn’t perceived as the actual problem.
They might think it’s ok to just use a loft or a sweep and have a huge tolerance of deviation and blame the network surface users as being the newbs
Oddly, my background began in the traditional parametric solid modeling programs before they had any good “free form surface” tools. So, back then for me, I approached things in the beginning treating everything like it was an “extrusion of a 2D sketch” solution for every problem – ‘cutting’ or ‘adding’ geometry and then filleting to finish everything.
Later, I needed to attempt to create ever more organic shapes. So, that lead me to using the now evermore famous “loft” 'ing approach. During those days I then thought lofting everything was the solution for every problem.
Eventually, I began using Rhino4 then Rhino5 everyday and learned how bad all the “lofts” really were that I imported from all those other programs I’d worked in.
I used to think “lofting” was the most amazing and most powerful methodolgy – until I discovered network surface and working with curve networks and manipulating the UV curves that are the underlying fabric of any nurbs surface.
When I became fluent in managing curve networks in the upwards of several hundred curves per situation, I learned that lofting is simply just childsplay.
Back in the “lofting is the solution to every problem days”, I used to spend weeks building sketches and guide-curves for a single lofting procedure that might have contained in the upwards of a dozen different sketches on different planes in order to attempt to remotely comprise some particular organic shape, and still fail to obtain 100% satisfactory results.
Nowadays, I would simply just network the surface, rough it out, and evolve it directly to become whatever I need it to be – directly composing the network until it is shaped exactly as it needs to be.
I surmise this might have something to do with degree 5 or higher UV maths, which I haven’t really ever tested on any CAM system.
I also don’t mess with altering ‘control point weights’. I simply don’t have confidence that any CAM system would interpret that data correctly.
I’ve been successful with deg 1,2,3 for almost 2 decades, and I guess I would have to begin testing those deg-5-theories to find out what I’m missing.
Until then, I don’t agree with what I’ve seen being said about deg 3 versus deg 5.
Yes, if a user is trying to match a deg 3 surface edge upto a deg 5 surface edge, then maybe they might think it’s a “to tolerance” thing, but to me that doesn’t mean anything new cause ‘tolerance’ is tolerance.
GD&T’s don’t care what degree your surfaces are, they still drive the outcome.
Therefore, deg 5’s etc are also a “to tolerance” thing and hence a false dichotomy to say they’re any less constrained by tolerances than deg 3.
Unless the relativity is to ignore tolerance, and let the deg 5 theory deviate blindly from where the design started and aimlessly discover wherever it goes – as long as it’s super smooth, who cares how much the geometry changes over time. As long as the final result is super duper smooth.
I would actually enjoy modeling things that way, but I’ve just never had the pleasure of such aimless freedom.
I’m very disciplined in abiding by the original geometries that I’ve received from any source and have always done my best to conform to the original shapes. Yes, usually these various shapes aren’t super duper smooth, and yes in the long run the design intent can always choose to do so, and go down the path of super duper smooth. But it’s upto the design intent. If a source says “yes deviate huge and make it super duper smooth” then so be it.
Thankfully Rhino has wonderful new tools and subd’s, and maybe degree 5 is the way to go, but this would have to be tested in CAM programs downstream. And I’m uncertain still in that regard.
In the end, one would prabably need CNC machines that can follow ‘spline curve’ data, which is a whole nother level I’m not sure I’ll ever reach.
For now I’ll have to use linear line segments and GD&T allocations, including file conversion translation degredation, tool marks, gouges, imperfections, polishing and some buffing maybe to strive for super duper smooth – in the real world results at least.
Cheers to ‘true fillets’ and ‘class-A’ surfacing 
