Gentlemen,

I am struggling with this apparently simple shape. I feel that there must be a way to generate a single smooth shape following these curves, but I failed to find it so far. Very frustrating !

Any idea?

One solution can be :)
You must first join lines broken!
After trim all lines between them
then network with 3 or 4 edges
and a sweep 2 rails when there are 5 edges by face (so join 2 lines for have 4 one)

Hi JPWEB,

Although it's not 100% perfect,here's my try with network.
The surface is not really smooth which is the result of
MOI mathematics (Michael can explain that better).

In your original file i found some curves not intersecting.

Hi JPBWEB,

This is what I was able to come up with using Network Curves. It's three surfaces. I know you want to maintain continuity, so this probably won't help that, but it is a way of producing that shape.

Hi Jean-Paul,

> I am struggling with this apparently simple shape.

Actually, shapes that kind of blend in many directions simultaneously and are not easily represented by a 2D profile curve are not really simple at all.

Often times things similar to that are more easily done in a polygon/sub-d modeling program rather than in NURBS.

But also you're basically trying to approach things in MoI with a strategy that is more oriented towards polygon modeling where you're trying to draw all the final edges of the resulting shape all in the beginning.

With NURBS modeling you should generally try to build some larger individual pieces, and then combine those together and fillet them to smooth things out between them - with this strategy you only draw a few original profiles and not a complete wireframe of all the final edges like you have done. Instead with the more typical NURBS strategy many of the edges will be the result of things like intersections and fillets and generated for you, rather than drawn explicitly at the beginning.


But as you get to shapes that are not well defined by 2D side profiles and are kind of looking like they are melting into different forms in multiple directions, that kind of stuff tends to be easier in sub-d or t-splines rather than in regular NURBS.

When you're trying to manually draw wireframes of things that look like they might be fillet or blend pieces, that's kind of an indication that things are not fitting well with MoI's workflow for that shape.

- Michael

Michael T's solution above is looking pretty good though.

When you have a small number of surfaces that meet in a crease and you want them to be smooth, you can cut away some of the area between them and use Blend to make a smooth connecting piece.

That looks like this:







So if those steps are done to Michael T's model above, it will basically fix the continuity issues there.

It looks like that may work well for this particular case.. But this is kind of moving in towards a much more difficult area of NURBS modeling and not really in the area where it is the strongest and most convenient. As your desired shape gets to be more kind of melted looking in many directions, that is usually an indication that sub-d is going to be easier for it.

- Michael

Thanks to everyone for the contribution. It is much appreciated.

Michael, your point is well taken. I appreciate that I am dragging MoI and NURBS modelling into an area in which it is not ideally suited. I insist on doing it though because I find Sub-D modelling anything but easy and because I am starting to get things out of MoI that really come close to what I want. I only post about the brick walls I crash into, but fortunately, I also get my way more often than not. What I like about the NURBS approach is the capacity to generate in a split second a huge and complex model made out of a few simple lines, as this is the way engineers created the real world planes and ships in the first place. In my mind, NURBS really HAS to be a better tool than Sub-D to make shapes that were intended to conform to the laws of hydro or aerodynamics.

I appreciate that it is often better to go for smaller components or for bigger pieces than can be trimmed as needed, and I am doing more of that now than in the past. The problem of that approach is to ensure continuity. A viable strategy for the tough cases might be to get as far as I can with MoI and then switch to Modo for the finishing touches. I really like MoI and I want to use it as much as I can to get what I want. So far it has been a journey of discovery and there have been times of huge frustration, but I get better at it, and this tips and findings from this very forum are a big help too. I should probably explain that my purpose is to build models of planes and ships that should be capable of 3D-printing. Overall accuracy of shapes is more important than minute detail and global prettiness. Hence the emphasis on continuity. This being said, it appears that I may have gone a bit overboard about that because the limitations of the 3D-printing process (although already black magic to my bewildered eyes) make it so that a slightly noticeable continuity flaw on my computer screen might be entirely absent in the real world output. As we continue to experiment, I should gain a better appreciation of the limitations. But better safe than sorry. If it is possible to generate reasonably flawless objects with careful planning and advanced NURBS modelling techniques, then let’s go for that.

Planes and ships, especially military ones, tend to be rather complex objects, but they are hard surface things. One should be able to get them modelled with NURBS reasonably accurately. An area that is causing me constant trouble is filleting/blending, hence the theoretical case I submitted to your collective sagacity. A real-world example of this would be the nose of a Lancaster heavy bomber as in the image below, but a submarine hull with protruding sonar would be very similar.


This is a very imperfect example and a rather old one at that. I got much better lately, like for the B-29 front wheel, a component way more suited to MoI's capability, I guess.

Here is a model with the blend features Michael G. showed.

You can create some "fairly" exact cap shapes with a method easy like this. (if you want to be "Precision", then you may have to practice a bit)

With this, I can get the same verticle cross sections as in your first posted filem which will produce a fairly close result.

Revolve a curve like this, being sure to create a point for each cross section you will need. Call this the resolution of your surface.



Then you'll get this result where you can delete the bottom surface and turn on points.



Each level of points can be moved up and down to match your cross section curves, and contain enough points to approxamate the shape that exists there. this will produce theis type of shape with very smooth transitions and curvature as you are looking for.



Sub-d NURBS.

[EDIT] Sorry, Seems I thought I was replying to this thread and shape:

http://moi3d.com/forum/index.php?webtag=MOI&msg=2909.1

But it may apply here also! [EDIT]

Indeed. I recently discovered, rather accidentally in fact, that one could manipulate surfaces using control points. This is potentially superb but a tad difficult to master. As it is with MoI, we can already use move, rotate, align with a cloud of control points, but it is often fiddly. I wish we could have tools to help with that, to complement what we already have, similar to Rhino 4.0’s Cage Editing capability if I understand well (I only have Rhino 3.0).

One great feature of MoI that I have come to use a lot is the way I can adjust a surface or object to fit blueprint drawings by playing with the control points of the curves that were used to build the surface, including with symmetry.

>>>>>>>One great feature of MoI that I have come to use a lot is the way I can adjust a surface or object to fit blueprint drawings by playing with the control points of the curves that were used to build the surface, including with symmetry.>>>>>>>

This is the way to do it BTW. :O

Hi Jean-Paul,

> What I like about the NURBS approach is the capacity to generate
> in a split second a huge and complex model made out of a few simple
> lines, as this is the way engineers created the real world planes and
> ships in the first place.

Yup, that's where NURBS is strong - that you can create a lot of shapes quickly by drawing just a few input curves, and using construction tools that use the curves.

With sub-d modeling you do have to kind of manage a lot of points floating around in 3D which is kind of a specialized skill. It's more spatially complex to work a big mass of 3D points than it is to work with 2D curves.

But the NURBS methods are best applied to shapes that are very well defined by their profile curves - when things are kind of melting/blending in many directions the profile of the shape is not really a primary definition of the shape anymore and that makes things more difficult.


> A viable strategy for the tough cases might be to get as far as I
> can with MoI and then switch to Modo for the finishing touches.

I think this is a good general strategy.


> Hence the emphasis on continuity. This being said, it appears that
> I may have gone a bit overboard about that because the limitations
> of the 3D-printing process (although already black magic to my
> bewildered eyes) make it so that a slightly noticeable continuity flaw
> on my computer screen might be entirely absent in the real world output.

Yes if your final output is going to be at a fairly small physical scale it would probably serve you well to kind of turn down the level of concern for getting perfect continuity.

If you don't worry about it as much, that will tend to free you up some more to build things in a kind of "skin the wireframe" type method like you were wanting to use here.


> Planes and ships, especially military ones, tend to be rather
> complex objects, but they are hard surface things.

Along with automobiles the exterior of those kinds of vehicles are kind of in a gray area where the shapes are becoming more in "organic" territory and less defined or driven by 2D plan curves alone.


> like for the B-29 front wheel, a component way more suited to
> MoI's capability, I guess.

Yes, definitely - that one is a perfect candidate for a profile curve driven approach.


- Michael

Hi Jean-Paul, I was messing around with your wireframe a bit more and I thought I'd share one note that may be generally useful to you when trying to use this kind of "make skin over wireframe" type method.

It's regarding this area where you may have a 3-sided region you want to fill in:




You've probably seen before where doing a Network on those 3 sided areas separately will make some undesirably bulging/puckering like this:



That happens because the curve pieces are of different lengths, and the upright profile can kind of sway back and forth as it gets merged with the other pieces, making for a kind of twist in shape.

But you can use Sweep with a special option enabled as an alternative which will work better in this situation, if you've got a simple planar profile piece like this:



What you do is select that planar profile, then run Construct / Sweep, and activate the "Maintain tangent" option:



That will cause the sweep to get generated by keeping that profile in the same plane and marching it down the rails and getting intersections with the rails and the plane, if you look at the control points of it you can see that regular planarity to them:



(EDIT: you would actually have seen it even more clearly if I had taken a top view snapshot, here to see that the surface control points are all arranged on parallel planes).

That's different than a regular sweep where the profiles kind of "slide" along the rails and can get that kind of twisting introduced if they are of different lengths.

Sweeps that are done with the "maintain tangent" option will more rigidly preserve the shape of the initial profile, which gets this kind of a result even when doing them as 4 separate sweeps:



- Michael

Michael,
Thats nice! One thing I was doing the other day when using network, was utilzing "RebuildCurve" to make all the curves have the same amount of points. I thought it was making a difference in the result! ???

Was I correct and does this relate to the different length thing you just pointed out?

Hi Burr, well it is hard to say anything for sure without seeing the particular curves...

But Network actually does a process like RebuildCurve internally. However, instead of doing it to a fixed number of points, it continues to add points until the rebuilt curve achieves a maximum distance tolerance away from the original input curves.

So if you smoothed out the curve or made it less wiggly or something by doing a RebuildCurve on it, that could possibly make the difference that you saw.


> Was I correct and does this relate to the different length thing you just pointed out?

Well, rebuilding several curves doesn't necessarily make them the same lengths even if they have the same number of points in them.

Maybe focusing on the length was not the right way to describe it - basically when you do a Network there is a kind of averaging process where the result is a combination of all the input curves in a equal way.

But in the kind of situation that I was showing previously, it is not really good for the input curves to be treated totally equally, it's better in that case for that upright profile curve to kind of have more emphasis placed on it kind of making it more of the primary shaping agent rather than a more equal blending between all the curves.

- Michael

Michael,

Thanks for your quality input. This thread has made me learn a few more tricks. I never quite understood before what the “Maintain Tangent” option did. I had tried tu use it but I saw no noticeable difference. Now I know (at least I think I know…)

On a related topic, what about this cage editing tool that is part of Rhino 4.0 ? Is this a kind of thing that would make smooth deformations easier for NURBS objects? Is it on the wish list for MoI ?

Hi Jean-Paul,

> I never quite understood before what the “Maintain Tangent”
> option did. I had tried tu use it but I saw no noticeable
> difference. Now I know (at least I think I know…)

Yeah it won't have an effect in just any sweep, the conditions need to be right for it to be able to be used, like the profile needs to be planar and if there is more than one profile they have to all share a same tangent direction where they touch the rails.

Also if you have exactly symmetrical rails then it would not really do anything because in that case the profile also slides along each rail in equal distances with each step.

It's when the 2 rails are different in shape and length that the profile will tend to twist as it is moved by sliding along the rails since the "slide distance" on each side is different in that case, except not with "Maintain tangent".

Maybe it should have been called "parallel planes" instead.


> Is this a kind of thing that would make smooth deformations
> easier for NURBS objects? Is it on the wish list for MoI ?

Yes, it would help to kind of squish objects around. It is on the wish list for MoI but I'm not sure when it will happen. It is difficult to make it work on a solid that has joined edges and still keep the edges stay connected in the deformed result, especially when the joined edges are trim edges that are located somewhere internal to the underlying surface rather than a natural surface edge.

- Michael