Well done Michael.

Export is deformed following the default. Clearly.

Hi ppj, for rendering purposes that will probably be ok though since the messy area is small in size and also kind of internal to the overall shape.

But that kind of self-intersecting surface will cause problems in many other kinds of operations, things like booleans will have problems operating in those areas because the intersection calculation between something like that is not very well defined. So from a kind of "NURBS solid model integrity" standpoint it is not good to have results like that.


Probably in v3 I will be able to experiment with an alternate mitering technique of basically taking the profile curve and rotating and stretching it and placing it at the corner areas, which should give a non-intersecting result usually but will probably not be 100% even tube thickness since it will make a kind of slanting effect.


- Michael

Hi Danny, it would be interesting if you could try the attached sweep example in NX and see what it does to it.

This one has a higher degree of non-planarity and bendy-ness to it, it should show how swoopy things are not really guaranteed to actually meet at a common normal miter.

I wonder if they will do something like twist the surfaces at the very end to force them to align or what...





- Michael

Hi Michael,

I knew you'd jump on this as soon as I posted it :) I was on my way out and was waiting for the wife to put on her finishing touches and didn't have time to analyse the sweep, I was going to try the tuber_miter as well but the wife had finished and we had to go :(

So I'm back now and posted the NX version of tuber_miter, and again no time to analyse it, but I'm sure you'll tear it apart Michael ;)

Cheers

Hi Danny, thanks for giving that one a try!

It's an interesting result - it looks like it is forcing the miter join by warping the ends of the sweep, I guess maybe in the same kind of tweaking of the control points at the end, although maybe it is a bit different than the other one.

That produces a kind of miter-join result but at the expense of not having fully regular thickness tube, just by eyeballing it you can see the thickness is different in these areas for example:





Here you can see why there is not a constant thickness - here I've set up a bunch of isoparm curves on one of the other surfaces and as you can see the warping at the end causes a kind of slanting to the cross-sections in the generated surface, they are not perpendicular to the central rail anymore:




However, that does not seem to really be a particularly bad result, because I'm pretty sure that it is a physical impossibility to maintain constant thickness in this situation and also have a nice matching miter, you can have either one or the other. And if you have bothered to have a single joined rail curve with a kink in it, I guess that probably means you are actually more interested in the miter.


- Michael

One thing I'm not quite sure what to make of with that NX result is that the shape at the miter here:



Has actually had a kind of shearing happen on it, if you lay that flat down on a plane it looks like this:




I guess it's actually a stretch of the profile which is something that goes along with a miter corner, but I'm not sure why it stretched in that particular direction.

- Michael

Hi Danny,

I am aware about the geometry kernel.
I want to know the result in NX, is NX capable or not ?

Thanks Danny :)

Hi Michael and Danny.....

I tried this case in SWX ( use the same kernel as NX ).
For case #1 = SWX cant produce any result. I dont know, it is the software or I dont understand the correct technique to get the result.
For case #2 = see attached

Hope this will give you more data to examine.

Thanks

Hi Anis,

> I want to know the result in NX, is NX capable or not ?

I've attached the results in the above posts, Michael explains his findings when examining the results from NX.

-

Seen your model on the Gallery, astonished result!
http://moi3d.com/gallery/viewitem.php?id=298
and seen you use the French UI ;)
Have you some remarks about it?

Hi Anis, well there's certainly no doubt that NX is a very capable program!

Of course with geometry there is such an infinite variety of cases that there will always be cases where some different kind of an approach is better suited for a particular situation.

That's why it is can be nice to have several programs in your toolbox when possible, because there just does not exist a single program that is always without question the absolute best in every single possible kind of geometric circumstance.


One nice thing about NX's result even in the one that it had a problem, is that it actually generated a result that can then be fixed up without much extra work by just deleting the short sections that have the problem and putting in something like Lofts in those areas.

It's not really that great when something just totally fails to generate any result like what you mentioned SolidWorks did for this particular case. That might be due to something in SolidWorks that detects that "bad geometry" was created and so therefore fails the whole operation. That kind of thing has sort of been a problem in parametric solid modelers since very early on in their existence.

- Michael

Hi Anis, also you wrote:

> I tried this case in SWX ( use the same kernel as NX ).
> <....>
> For case #2 = see attached

It's interesting that it is such a different result - it must be using some other variation of sweep or different settings or something like despite having the same kernel.

This result is really pretty bad with such bad wiggles and twisting in the shape, and it actually also has a kind of severe folding back on itself type problem right at the miter location as well.

- Michael

So the key is not only engine of the software.
But also who develop the software use the engine, right ?

Thanks Michael for your valuable info....

Hi Anis,

> So the key is not only engine of the software.
> But also who develop the software use the engine, right ?

Yup, that's definitely true.

MoI is an example of that as well, I re-wrote a large amount (like maybe 60% or so of the total or maybe even more) of the sweep code for MoI's sweep function from the stock one in the geometry library because it had various restrictions that I didn't want to have (stuff like profiles must touch the rail curve, etc...).

But the custom code that I wrote still uses a lot of functions from the geometry library for various low level calculations as well, so the custom code still depends on the library in a lot of ways.

You may find that same kind of situation in other programs as well, where some programs may have developed some functions differently from one another despite having the same base geometry library.

- Michael

So.... going back to his original question about the sweep, is there maybe an alternate method to achieve the desired result without using another program?

I think you need to create profile in each corner

Hi niko yeah currently without using any other program you need to get angled and scaled profiles into the corner junctures, then build sweeps between each of those.

A shortcut for getting the angled and scaled profiles is to make a polyline for the rail - if the rail is only made up of line segments then you can get a mitered corner sweep for it in MoI even if the whole rail is not on one plane.

So for example try sweeping the attached 3DM file in MoI - here I've drawn in a polyline for the rail instead of the original one that had slightly curved segments.

That produces a result like this directly in MoI:




If straight sections are ok then you are done there actually.

But if you want curvy sections, you will then want to extract the edges at the polyline miters by selecting them and doing Copy + Paste, and then do a sweep on each curved segment individually, between those profiles.

- Michael

Hi niko, here's an example of how to use that technique with a curved rail, this is how to build the second example I posted entirely in MoI.

I started by drawing a polyline snapping its points on to the rail like this:



Now sweep along the polyline, it will build a mitered result like this:



Select just the edges around the original juncture point, do a Ctrl+C and then Ctrl+V to duplicate the edges as regular curve objects, and then delete the segmented sweep, giving you this:




Now use the Transform > Orient command to copy the starting rectangle to the very end, or actually you could do it by duplicating those edges from the sweep as well:




Now it is ready to sweep, use Edit > Separate to break the rail into individual pieces, and do the sweep in sections, first one sweep here:



Then a second one here:




That's how I produced the attached 3DM file created only in MoI.

- Michael

Hi Michael,

>>That's how I produced the attached 3DM file created only in MoI.<<

I presume that result is just for fun, as it is a very bad output.
You will not produce a good result attempting a single angular bisector/trim at the 3d intersection

Hi Steve, sorry I don't have any idea what you are referring to - that's not a bad output at all, it's a full solid and seems to check out fine, no self intersections, etc...

Also I don't understand why you are talking about an angular trim - the sweeps in this case touch each other exactly because they come to a common angled profile.

That's the whole point of this method - to make a joined touching piece without using any trimming.

You have to use some method like this to make pieces join in this kind of situation because 2 regular shaped tubes will not really come together in a simple miter like that when the paths are all curving around in 3D space, that's what I've been describing in several previous posts in this thread with several examples.


Using an angled profile shape that is common to both sweeps, does make a result that touches, at the expense of it being a slightly slanted tube and not completely even thickness throughout.

If you think it is a bad result, you will need to explain what your complaint about it is - as far as I can tell it's all fine, for example here I have done a boolean in that area with a sphere and as you can see it gives a clean result:





- Michael