The thing is that if one breaks the curves, then the continuity of the hull deteriorates greatly. See this zebra analysis of your model:



A ship hull is a sort of semi-organic shape. It should flow in a nice fair uninterrupted curvature. Accuracy of individual stations is less important than maintaining fairness. It is a perennial problem of mine with NURBS modeling (not that it would be any better with Sub-D obviously).

Hi JPBWEB,

Something more like this?

Michael T.

This is more like it. I suppose you went for two hull halves but there is a gap (and thus an naked edge) between the two hull halves. If you delete one (they are slightly different btw, I wonder why) and mirror the other one, you can join the two with no naked edge but there is still a detectable kink. In this particular case it would be good enough because there needs to be a longitudinal keelpiece added that would hide it, but still this is why I was trying to generate the entire hull in one go.

It seems possible to get much closer to the stem line with a V-shaped frame and still get the network curves to work, but there remains a gap at the pointy stem. This can be filled in several ways, but it seems hard to ensure continuity. Argh!

Actually, I spoke to fast. I used a simple Blend on the edges of the v-shaped hull, which gives a perfect continuity. Varying teh bulge value (within reason), one can get various shapes of pointyness (if that is the right word).


Here is a Flamingo rendering showing a very smooth stem-piece. The zebra analysis shows perfect continuity.


This means problem solved, I guess. At least for such a simple smooth hull form.

Hi Jean-Paul, it looks like you have got this all figured out now, but just to clarify your original problem, that original layout you had did not form a completely regular UV grid with all the curves, which is what Network needs.

For example this profile touches the outside one at 2 points:



But then the next one only touches the same outside one at only 1 point:



You can also kind of see there that there is a difference in length - that last one only matches up to something like half the size of the previous one, it's not fitting into a regular grid type formation.


Another thing is in Network when you have one direction collapse down and touch each other like this:



It can't be just a partial amount of things that collapse, all of the curves for that direction must collapse to that same point to form a complete "pole" there.

Basically Network needs to remap your curve structure into a completely regular 2D rectangular grid in UV space, and if it can't figure it out it will fail which is what was happening to you originally here.

But with the updated layout where you have a slight gap in the front, that then gives Network the kind of structure that it needs.


Maybe in the future at some point I could make Network more intelligent and have it figure out your original layout, but it would basically need to recognize that end curve there is a special case and have it internally be doubled back on itself to form the complete UV grid. But it may be fairly difficult to set that up.

- Michael