Today i tried this algorithm of patch creation and have problems with naked edges again.
Guess i missed something important in trimming process or in surface creation.
I assume that one of a steps should guarantee the right contact of result surfaces.
I just cant figure which one. I inspect the project you provide Michael. And it works like a charm.
So i recorded short video of whole process.
https://www.dropbox.com/s/8nms9s7buihvk56/2019-02-03_12-16-25.mp4?dl=0

I also add project file to inspect if necessary.

Hi Bravlin, I took a quick look at your 3DM file and it looks like you're doing a Loft instead of an Extrude perhaps?

The curves you are lofting between have some skew between them, so that's going to affect the accuracy.

I could see the skew by looking at this edge in the Top view, a completely vertical extrude should only show as a point in the top view:





And it looks like the top and bottom curves don't have aligned endpoints:




That method I showed previously is only applicable to reverse engineering a completely vertical extrusion surface, not one lofted between 2 slightly different curves.

- Michael

> That method I showed previously is only applicable to reverse engineering a completely vertical extrusion surface, not one lofted between 2 slightly different curves.
Yes. That's exactly why i use loft between two curve profiles.
I noticed that upper and lower edge silhouettes lie not on completely vertical line.
Can we solve such case without patch tool (inventor, rhino) ? If its not completely vertical ?

Hi Bravlin, sorry no you can't accurately build a surface that passes through several interior non-isoparam curves like that, other than the full vertical alignment case.

The main way to do it would be to approach it how you would have wanted to model it originally which would be to have your pieces here as an extended base that has flat ends:



Then you would make your bendy front face as an extended surface shaped how you want and then trim or boolean difference your base piece with that forming new edges where things intersect each other.

You want the edges you're trying to fill to be formed by surface/surface intersections, not by keeping them and trying to build a new surface that has to pass exactly or very closely to all of them at the same time. You can do that if all the edges you have to pass through are a 4 sided region so you could use Network but not when they are in the middle of the surface.

- Michael

Hi Bravlin,
I un-trimmed your surfaces and found a problem wilt the 2 faces shown in the photo they do not extend far enough to be trimmed by the vertical planar. Anyway having overcome that I made a solid of the body (no cutaways) and a solid to Boolean subtract (with cutaways)
to get the finished Solid.
The difference between top and bottom curves is 0.19 don't know how important maintaining this is ?
I generated the curves similar to how you have done but then extended them by using the extend tool.
This allowed spare for trimming.
Hope this helps
Barry

Thank you for the effort Barry-H.
My question is more about approach and limitation i guess.
To better know the boundaries of a program and algorithms of work.
We established that secure patch creation limited by one plane without tilt.
For other cases we shall use other software (solid, rhino etc). Which have patch tool but also
have there own limitation by form and curvature of result surface.
Also thanks for reminder about negative shape boolean approach.

Hi Bravlin.
attached for reference purposes is the solid I created which maintains the 0.19 tilt.
Cheers
Barry

I tried reproduce Barry-H negative boolean approach.
But already have one additional question.
I dunno how does you reconstruct such a clean boolean shape(blue layer shape).
If we select naked edges and then join them we would have pretty messy curve.
And if we use rebuild curve operator even with Tolerance: 0.001 we would lose straight part of curve.
So i guess that this boolean shape was constructed from original curves and surfaces (light ones).
So, can you please share, how did you do that ?

Hi Bravlin,
I select the edges from the flat faces each end I want to use copy them to clip board and paste back thus giving me curves not edges
that can be used to make the Boolean shape.

Cheers
Barry

Its strange. Ctrl+C, Ctrl+V creates pretty messy and dense curves. It also creates lots of them not the one closed curve.

Hi Bravlin,
Select these edges each end copy paste and join.
You then need to extend the lines I showed in the previous post.
Join with lines to complete the Boolean profile.
Barry

Hi Bravlin,

re:
> Its strange. Ctrl+C, Ctrl+V creates pretty messy and dense curves.

Often times curves from intersections can be pretty dense. It's normal and not usually a problem since you won't normally be control point editing such curves.


> It also creates lots of them not the one closed curve.

You'll get one curve for every edge that was selected. Alternatively you can directly use Edit > Join on multiple selected touching edges to make a closed curve result without doing Ctrl+C/Ctrl+V first.

- Michael

Frenchy Pilou (PILOU), thanks for your advice about sphere.
Today i stuck with rail and profile approaches and your advice really helps.


The result so far looks like this.


And the profile try looks like this one. I'm still bad at cad surface creation.

So if someone wants to share its own try it would be great.
This part is well known M. A. Nash mech (hope he never know)

For this sort of thing only patience is necessary! ;)

By the way, what's wrong with that curve ?
When i add point it jumps away.
https://www.dropbox.com/s/ez69mpi45g985m2/WJW0Sy3GRX.mp4?dl=0

Hi Bravlin, that looks like a bug, I will investigate.

You can add a point by knot insertion instead of control point insertion though. That's when you click directly on the curve instead of on the control polygon. It's easiest to do that if you have control points turned off when you are in Add pt and then click on the curve. That will refine the curve in that general area without changing the shape of the curve.

It's normal when adding a control point to the control polygon that the shape of the curve will change a bit but the new control point is not supposed to shoot off like that.

- Michael

Hi Bravlin, the bug has to do with the curve being a rational curve which means it has additional weight values in addition to the x,y,z coordinates for every control point. Usually only exact conic section curves like arcs, circles, and ellipses use weights. This one probably has them from combining an exact arc with a neighboring curve.

I'll see if I can get it to behave better.

- Michael

Hi Bravlin, yes it was a bug in handling inserting control points into a rational curve that has weights. It's fixed for the next v4 beta, thanks for reporting it.

- Michael

Q_03: Hope it's not too much if i ask 2 questions one day. Got another shape-question.
We discussed earlier the network-operator approach. And for smooth shape it works great.
But today i tried to use this approach to construct this shape. This shape a bit more complex but still have rounded sides.
I spend half of a day in attempts to create clean shape and generate plenty of guesses in a process.
So if it possible can you share some guides rules to network operator according this shape.

I attached project file that have 2 layers.
On layer "L_09" i trimmed profile curves in direction of middle line, and on layer "L_10" i trimmed it in axis direction.
But i guess since we have shape that do not have equal sides its not an issue.




In result we should have something similar to that shape.

Hi Bravlin, sorry I'm not fully understanding your Network question. Is it that you would like it to make a shape as if it had already filleted edges in one single application of Network?

That will be difficult since Network builds one single surface result (which will be split up at any sharp corners) and it is difficult to make a single surface that goes through a localized significant change in shape like a fillet transition.

In general it's best to use Network to build one smooth surface which has a regular 4 sided boundary. It isn't generally very workable to try and build what should be a multi surface model out of one single Network.

Over time I have noticed a kind of recurring pattern that people with a poly modeling background really like to draw all the 3D edges of their model and then want to use Network to fill it in. But with NURBS modeling usually it is better to make extended pieces that do not have pressure applied to their boundaries, that's how you get high quality surfaces. Then some of your final edges come from intersections like trims or booleans rather than trying to do it all with directly constructing surfaces to an irregular boundary.

- Michael