ugly creases in network

Next
 From:  niko (NICKP100)
4844.1 
I do not understand why I'm getting that creasing towards the tip of the hull.
http://img856.imageshack.us/img856/9150/creases.jpg

When solving the exact same network in rhino v5 the result is clean!!

EDITED: 10 Jan 2012 by NICKP100

Attachments:

  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
4844.2 In reply to 4844.1 
Hi niko, one thing that is contributing to that is that there is an actual crease in one of your curves here:



That segment at the end of the curve there is not smooth with the other piece, their end tangents are about 6 degrees different from one another, so there is a kink in that curve there.

When Rhino builds a network, it basically ignores any kind of kink in any of the input curves, which happens to work to your advantage in this particular situation but can cause errors in other situations where the kinks are not respected.

In order to get a better quality network from that in MoI you need to make your long curve there be of better quality and not be made up of separate segments with actual sharp points between the pieces. That would probably help to some extent anyway.

It looks like you may be getting a little bit of wiggling in the curve refitting mechanism as well in this particular case - I'm hoping to do some work in v3 to try and reduce that.

- Michael
Attachments:

  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Frenchy Pilou (PILOU)
4844.3 In reply to 4844.2 
< to try and reduce that
By an alert message or an automatic modify ?
---
Pilou
Is beautiful that please without concept!
My Gallery
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
4844.4 In reply to 4844.3 
Hi Pilou,

> By an alert message or an automatic modify ?

By drawing a new curve to replace the old one or possibly using the Rebuild command to reconstruct it. For Rebuild I think the angle of the tangent deviation is enough that the rebuild to tolerance mode will see it as a sharp kink and rebuild it in segments, but the rebuild by # of points option can be used instead. However, it may be better to just redo the entire curve.

- Michael
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  niko (NICKP100)
4844.5 In reply to 4844.4 
Well I redrew the curve with less points but I still get the same result.
The reason I put the points close together at the "kink" that you pointed out is that there is no other way to define that tight corner by spacing points further out.
Any ideas?
Attachments:

  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Mauro (M-DYNAMICS)
4844.6 In reply to 4844.5 
Hi Niko:
i downloaded your file and i found two end points not matching other curves (yellow arrows)





after moved them i did a sweep with red curves like section and two blues like rails
have a try...
i also did 3 network surfaces without moving your points,just trunk them
take a look at the file i attached
hope i understand what you want or maybe you need the corner more "sharpened"?
M-D

EDITED: 11 Jan 2012 by M-DYNAMICS


  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
4844.7 In reply to 4844.5 
Hi niko,

> Well I redrew the curve with less points but I still
> get the same result.

Actually if you look at the first version you posted and then the second version at the same time I can see some improvement in the second version.

A few other things you can tune up are as Mauro mentions above some of the curves do not actually touch each other, and some of them are a little wobbly instead of being planar.

I tuned up those things and got a little more improvement from that, and then sometimes with network if you have some bad shaping near the tip where things are collapsing down together it can help to put in an additional section pretty close to the end to help control the shaping just before it all collapses down.

I've attached the 3DM file here, those adjustments seem to give a better result:




The other thing that you could try is to avoid putting in the tightly bent area directly into the network and build something that's more like 2 separate sections and then either fillet or blend to make the tightly bent portion - having that somewhat irregular distribution of curvature in the one profile can tend to kind of ripple out when that portion is trying to get blended with the other profiles which are not shaped in the same way.

- Michael

  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  niko (NICKP100)
4844.8 In reply to 4844.7 
I hate to say this but looking at the mesh and polygonal export I can still see some creases (admitedly a lot less than before) at the front of the model and where that small curve is....besides I do not think that we are always in the ideal situation of being able to have all planar curves

In Rhino, comparing the wireframes side by side I can also see how much cleaner the rhino solution is. IMO the sharp edges still show as expected and the surface is smooth where its suppose to be. I cannot imagine an example where the MOI solution would be more favorable...

The Rhino surface is to the right while the MOI is on the left. The MOI surface is more dense but the Rhino one is a lot more uniform thus generating less unexpected results

Michael I was wondering if there is a chance that you can give us a choice on how to solve networks in the future.
Image Attachments:
Size: 148.9 KB, Downloaded: 59 times, Dimensions: 912x556px
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
4844.9 In reply to 4844.8 
Hi Niko, yup I wrote some about that previously:

quote:
It looks like you may be getting a little bit of wiggling in the curve refitting mechanism as well in this particular case - I'm hoping to do some work in v3 to try and reduce that.


It seems to be a little bit of wiggling that's happening in the curve refitting mechanism when it's processing some of your particular shapes here.

Since Rhino's fitter is working better for this particular curve network that you've got here, why not just use Rhino for building this particular Network instead of MoI, wouldn't that solve your problem?

- Michael
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
4844.10 In reply to 4844.8 
Hi Niko,

> In Rhino, comparing the wireframes side by side I can also see
> how much cleaner the rhino solution is. IMO the sharp edges still
> show as expected and the surface is smooth where its suppose
> to be. I cannot imagine an example where the MOI solution would
> be more favorable...

I've attached here a simple example where the Rhino solution is not favorable, if you select all the curves and run Network MoI respects the creases and generates the proper output:



Take particular note of the corner areas here:





Now load the 3DM file into Rhino and construct the network there by selecting all the curves and running NetworkSrf. If you examine the same corner areas:





You can see there where the network does not adhere to the corner shape, it makes a smooth surface through there instead of actually coming to a sharp corner. That means that for example if you tried to extrude the bottom shape downwards and then join the extrusion and the upper part, they won't join together properly because the top part kind of bulges away from the sharp corner.

You get similar problems in Rhino trying to use any kind of creased curve to build a big network.

- Michael

  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  niko (NICKP100)
4844.11 In reply to 4844.9 
Because I love MOI so much that I want it to be my exclusive nurbs modeler .

I don't know what wiggling is, but will that particular network solve closer to Rhino after you work on what you said?

Thanx
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
4844.12 In reply to 4844.8 
Hi niko, attached here are few other examples to help explain the problem with just trying to ignore creases, try loading these 3DM files into MoI and into Rhino and do NetworkSrf in each.

In MoI you get this result:





And in Rhino take a look at what happens to the corner areas:






Anyway, I hope that gives you some better idea about that issue.


- Michael

  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
4844.13 In reply to 4844.11 
Hi Niko,

> I don't know what wiggling is, but will that particular
> network solve closer to Rhino after you work on what
> you said?

I hope so, but it's hard to know for sure until after it's actually done.

You also might try using MoI v2 for this one instead of the v3 beta - one change in v3 is that it uses a tighter fitting tolerance for rebuilding the curves, but that can have some side effects like this.

That's another area that I want to work on during v3 - to be able to have a tighter fitting tolerance only for the outside curves of the network that might need to be joined to other things, and a slightly looser tolerance (which makes a lighter density surface) for the internal curves.

But anyway if you try it in MoI v2 you should probably see a lighter surface generated.

- Michael
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
 From:  niko (NICKP100)
4844.14 In reply to 4844.13 
Yes it is lighter and i much much prefer it over v3, but it still generates non-uniform spans and still looks pinched where the surfaces come together . Hopefully your solution will work for V3....

EDITED: 12 Jan 2012 by NICKP100

  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged
 

Reply to All Reply to All