Surfacing A Corner (Should Be Simple)

[HeathBar]

I've been a long-time lurker on these forums, and have learned a LOT from you all over the years. I've come across this problem a number of times and have thoroughly stumped the other designers I work with, so thought I'd query The Brain Trust.

I have a pretty simple corner between three 90-degree planar faces with three G2 Styled Blends (5-degree single-patch) and a G2 Bridge Curve (modeled as a Law Extension so I could export a parasolid). I would like to fill that hole with another G2 surface.

Mesh Surface and Studio Surface both produce high patch-count surfaces that bulge out and cause issues trimming the planar-solid. The Sweep tools don't provide any tangent or curvature continuity. My best effort yet has been to create the single-patch using a G0 Mesh Surface, use Match Edge to control the continuity to "the corner" Blend, and use X-Form to manually move poles into the correct position. Here my personal favorite tip comes into play; sticking a post-it note to my monitor as a straight edge.

Alternately I can extract Point Sets using B-Surface poles, create Lines and Planes, and manually create the surface I want using Surface From Poles. This creates the perfect version of the surface above, but it's non-associative which goes against our modeling best-practice.

So the question is; is there an associative tool which will create this corner, G2, single-patch?

 

[TheTick]

Not sure why you're concerned about patch count.

I tried Studio Surface with c2 continuity all around and result looks fine. Pole density is a bit thick but commensurate with input.

 

[HeathBar]

Normally I'm not overly concerned with patch count (within reason), what really drove this issue was that my Studio Surface/Mesh Surface was pulling out from the original curves which is causing a Trim failure.

Where patch count comes in; it is possible to make this a single-patch feature. I'm just having a hard time stomaching that NX doesn't have a feature to create this surface. And I'm hoping I'll learn something along the way.

 

[Toost]

I assume you have played with the "Styled Corner" and the "Blend Corner" features ?
I have not tried your model on this. I cannot say whether those will do a single patch solution or not.
I assume that a single patch 5°x 5° G2x4 surface is mathematically a pretty tricky thing depending on the surrounding surfaces, in this case it is possible but with very small deviations it isn't and then the modeler will need use multiple patches and tolerances.
What NX version ?

Regards,
Tomas

Never try to teach a pig to sing. I wastes your time and it annoys the pig.

 

[cowski]

For fun, I tried "fill surface" on your part. I used G2 for the large surfaces and G1 for the small surface (it wouldn't take G2, but since it is a planar surface, I didn't think it would matter much). The reflection lines on the resulting surface looked surprisingly good. I didn't check patches/pole count/etc.

FWIW, a couple of my general observations on surfacing in NX:
[ol 1]
[li]If you want something associative/easy updates, you generally use a common NX command and just take what it gives you. If you need a "perfect" surface, you build it manually and have to tweak it almost every time the model changes.[/li]
[li]Me: "this should be simple"
Narrator voice: "it wasn't"[/li]
[/ol]

www.nxjournaling.com

 

[HeathBar]

Thanks for the tips, all. I'm using NX 1867. It might also be useful to mention that I have my tolerances cranked way up, 0.0004 for Distance and 0.01 for Angle (ips units) based on company surfacing best practice yadda yadda.

Styled Corner is a really good suggestion, I've used Blend Corner before but haven't tried Styled Corner. I like it a lot and will probably come back to it. Unfortunately it bulges out in the middle and causes a trim failure, same as Mesh Surface and Studio Surface. Fill Surface produces the surface closest to what I'm looking for; doesn't bulge, pretty low patch count.

It sounds like everyone is equally stumped regarding the creation of the single-patch surface. Like I mentioned, I've come across this a couple of times. I think NX is programmed to maybe overlook the elegant solution and follow the solution that is reliable in all situations. What I mean; two intersecting curves/lines can be connected with Bridge Curve at G2 with a single-segment, 4-degree, 5-pole spline. Two poles are at the endpoints, two poles control tangency, curvature is controlled at the intersection point of the two lines. But functionally NX will always create either a 2-segment 4-degree spline or a 1-segment 5-degree spline, presumably because that function won't fail for non-intersecting curves/lines.

I feel like NX follows similar functionality with the surface tools as well; it likes to create at least one extra row of poles to de-couple Surface 1 from Surface 2 so that tangent/curvature can be independently controlled on each side and presumably so the tool works in more situations. I understand *why* it does this, I just find it interesting that step 1 is to overlook the simple solution. I feel like instead of being programmed to look for the elegant solution NX just takes a sheet and tries to stretch it around until it meets the boundary conditions within tolerance.

So thanks for the help all! I'll consider this satisfactorily unresolved. If someone with GTAC or an NX programmer wants to weigh in I'd be happy to continue the conversation but for now I think cowski nailed it; "if you need a perfect surface you build it manually".

 

[TheTick]

Why are you so hung up on a single patch?

 

[Toost]

Single patch is the holy grail in Class A surfacing. If you have single patch and a degree which matches the task at hand, you can sleep well when the tooling is about to be cut.
There are no, there cannot be any, shape defects such as ripple, wobble, bulges, bumps etc if the pole structure and the amount of poles+ degrees are correct. This simplicity in math is the guarantee for the shape quality.
If it is needed in all instances is a different question. The surfacing gurus tend to have difficulties not to create this level ...

Regards,
Tomas



Never try to teach a pig to sing. I wastes your time and it annoys the pig.

 

[TheTick]

You're trying too hard to achieve something that helps no one and is somehow making things even worse. You are wasting a lot of time. I made a good surface in less than 10 minutes from download to upload.

Good surfaces are good surfaces. Continuity is all you need.

Don't tell me about surfacing gurus. I will tell you.

 

[mmauldin]

I threw a few extra guides, look a little better (to me).

 

[JohnRBaker]

Someone said that they tried the 'Fill Surface' function. Did it work? I mean this guy makes it look easy:

https://www.youtube.com/watch?v=0Jx81mrmm_c

John R. Baker, P.E. (ret)
Irvine, CA
Siemens PLM:
UG/NX Museum:

The secret of life is not finding someone to live with
It's finding someone you can't live without

 

[HeathBar]

I aim for low patch-count for all the reasons Toost mentioned and also because patch-count tends to increase as each additional feature is created. If I start a design with a multi-patch surface and I need to use that surface to create another surface, I can guarantee the second surface will be higher patch-count than the first. If you don't pay attention you eventually get to the point where you're actually generating noise at a low-level and, like Toost said, you end up with little ripples that aren't intended. On a bad day the continuity of your surfaces begins to rely on the modeling tolerance and you start getting those "cannot comply with tolerance" errors. Practically many patches might not be a huge issue because it's all on the 1/25,000 in level. Realistically, since the vast majority of my work gets Wave linked into other peoples' parts, I'm going to end up with someone's manager standing at MY desk wanting to know why his guy can't make simple model features work on his simple part.

All that said, knowing where this particular corner is in my feature tree I'd be completely satisfied with up to a 10-patch x 10-patch surface. The exercise here is academic. *I* can make a single-patch surface here, why won't NX?

John, that guy (whoever he is) sure made a good pitch for fill-surface. You use it in exactly the same way I do; filling in holes in Industrial Design data. It also is a good way to make the pesky 3-sided surfaces IMO. In this case it does create a surface which fits inside my boundary data but the pole-count is a bit of a mess. Recall, my Sew tolerance is really high.

Here is my most frustrating attempt to date. This is Studio Surface (although Mesh Surface looks very similar) with G1 and rebuild as Auto Fit 5x1. *SO* close...