Hi Burr,

> We've discussed this before Michael, with regard to Rhino
> and you described your thoughts that having that variable
> tolerance value was a design mistake that lead to these
> types of situations.

Yeah, what can happen is if someone fiddles with the tolerance it can easily lead to geometry being too saggy or it's also bad for someone to make it super tight and get things like 100,000 control points on every single trim curve.

It's especially surprising when you receive a file from someone else when they have messed with the tolerance and you are not aware of it, and then you start to create new things after loading in their file, getting things like sags in your fillets.

- Michael

Hi Burr....

BTW I dont receive yet any request from you.
PM me and provide your email if you want....

Thanks

Hi Michael,
It must be a saggy thing with regard to the model I'm refering to. It is a simple revolved solid with holes punched in it and fillets on the holes. Nothing manually created. You would think that the fillets would create a nice edge as opposed to naked. I do suppose you could still punch holes that would cause it to be faulty fillets too though... Many factors there.


Anyway, mine is not a MoI model, so I will wait for Anis's to see if I can conjure up anything....

Thanks again.

Hi Burr, does the one you have join up if you scale it down by 1/10 or 1/2 using the steps from here:
http://moi3d.com/forum/index.php?webtag=MOI&msg=3593.3

Possibly that model was built to a tolerance of 0.01 units, maybe making the fillets just sagging away by slightly more than MoI's max join tolerance. A sag of 0.01 units is not particularly visually bad, it can be tough to see it unless you examine things more closely.

The reason why scaling down can make things join up is that when you scale an object down, the gaps between pieces also get smaller as well, so it can put them within the tolerance distance if they were just outside of it before.

In the future I would like to make a way to override MoI's tolerance, but probably something like an override that lasts just for the current command or modeling session rather than something that gets stored in the file and set once the file is loaded - that particular method of having a changed tolerance get set on someone when they don't expect it is the thing that tends to cause trouble.

- Michael

Hi Michael,
I did try the scale and it would join more than without the scale, but never make a solid.

This is the model I refer to. It is not something I need to fix. It is being used for a question regarding machining....But it just seemed to be the same thing so thats why I mentioned it. If you want to peek and refer to it to see if anything stands out, that would be nice. But there is no need to "work it" to try and make a solid. Only to see if you can see why it is not.

Thanks

Hi Burr, I'm not 100% sure what is happening there, the edges do not seem to sag away from one another, but also the edges between different pieces are not exactly aligned with one another, some of them have different endpoints although not very far away.

I took a close look into one problem area and it looks like there is some crossing overlap between some of the edge curves, that may be the problem.

Like for here I marked the end of this edge with a point:



Now if I go select the edge right next to it you can see this:




I'm not entirely sure if that is the problem though.

- Michael

Hi Anis,

I had a look at this model and unfortunately it's a mess, it seems that the person who modeled this part has had the modeling tolerance set very loose in their cad program.
Ralf was correct, it was exported from NX, I've highlighted it in the picture attached.

I did run some analysis in NX, also NX has a feature called heal geometry, what that does is it tries to fix geometry by tightening tolerances, removing tiny objects and various other surface repairs, it works quite well when there a few geometry abnormalities, I've also attached the output result of that operation and you can see how many things are wrong with this model, but there were still a lot of naked edges after this operation, the best you could get was to sew(join) the surfaces in NX using a tolerance of 0.1mm(0.004") which in my opinion is way too loose, I wouldn't trust the quality of any of the surfaces and it still resulted in a lot of naked edges that would take some hours to fix.

Anyhow there is no quick fix for this file, I suggest if your client is using NX 5.0 I can't see why they can't output it as a STEP to see if that is any better.






-

Hello,

This is the basic result from an import to SCE and a stitching command. I think it almost might be better for you to try to defeature the model and build it back up. Even it its somewhat whole state it has hundreds of geometry errors. Is this for rendering purposes?

OSTexo, was it hard job to stich the surfaces with SC?

""""""""""" but also the edges between different pieces are not exactly aligned with one another, some of them have different endpoints although not very far away.""""""""""""""

Thanks for looking at that Michael. I thought I inspected those endpoints of edges and didnt see any. I need more practice!

Anyway, it seems that with this, and Anis's model, it's not really what I was looking for, but more construction issues, or the export didnt handle the solid creation well. "STEP" up to solids :o

Thanks again.

Hello d3print,
It wasn't difficult at all. I just selected the entire lump of geometry and hit the button. I checked the geometry of the model and even after that there are a bunch of problems. I have a little free time and am trying to use it as a challenge to sharpen my skills. It may not turn out anyway but at least I will get some practice. It seems the edges are sort of sloppy and gapped, overlapping, etc. in a few areas among other things and the rounds look messed up in many places. SCE has some great defeaturing tools so once I hopefully get this into a solid I'll be able to tighten that up. I'll try to isolate the pieces and parts from the largest mass of the model to indicate what surfaces are causing problems.

Something that was built to a loose tolerance plus at the same time has small sized edges or features in it can tend to be problematic.

- Michael

Hi Burr,

> "STEP" up to solids

Yeah that can often help, particularly with not having to join things.

Unfortunately STEP has some totally different areas where things can go wrong in - the way trimming boundaries are specified in STEP files, only the 3D curves of the edges are contained in the file, leaving the UV space trim curves to be calculated by the application. That calculation can be fairly difficult in some circumstances, particularly in seam areas of closed surfaces if the trim curve wanders slightly back and forth over the seam.

Also the STEP mechanism allows for things like a single trim curve that crosses over a seam in a closed surface which has to be split for regular NURBS topology. That can again be a difficult calculation if the edge just barely grazes the seam or hugs along it for just a little bit instead of crossing directly over it, stuff like that.

So there are some trim curve calculations and processing that still has to happen with STEP files in order to get it into a normal NURBS topology, and along with any kinds of fairly complex calculations there is potential for things to go wrong in grazing / barely touching / overlapping kinds of situations.

It sure would have been nice if the STEP authors would have made it a more direct container of simple NURBS topology, that would have made it a more ironclad data transfer mechanism.

- Michael

Hi Burr, so I was experimenting some with different internal options for surface joining, and by turning on an option for squeezing together small edges I was actually able to get your wheel thing there joined into a solid (just 1 surface needed to be trimmed), it is attached here.

But unfortunately I have also seen that "squeeze small edges" option cause some kind of orientation problems in some other cases, that's why it is not turned on currently in MoI.

That seems to be the actual problem with the model though, not that the endpoints have that little overlap, but that there are some little super tiny micro edges in the trimming boundaries, actually at a 90 degree turn from their surrounding edges making for a really messy situation.

Here I have tracked down just one of them:

In this area here:



If you zoom in you can see the 2 larger edges have a kind of criss-cross arrangement (like I was showing earlier, this is another angle on the same kind of thing):



which is not really great compared to touching nicely, but actually the ends of the criss-cross are within 0.0007 units of each other which is within tolerance so that is actually by itself not going to interfere with joining.

But notice the circled area, there is a little spec of yellow in there - that's actually another edge in there. Here i have zoomed in more closely on it:




And here it is unselected so you can see it more clearly:




So that's a terrible super tiny little edge hanging off one of those larger edges, and it is at some kind of 90 degree angle to the overall boundary shape, just really nasty to deal with trying to join that to something else.

That little micro edge is only 0.0000007 units long - this becomes bad when the entire edge is way smaller than the tolerance, it becomes difficult for the joiner to figure out how its endpoints should relate to other way larger edges that it is trying to join it with.

So anyway, that's what's going on with your example - it's not sags, but rather badly formed trimming boundaries with tiny micro edges in them.

At some point I could probably cook up some kind of "edge repair tool" that would do things like try and just remove little tiny micro edges like that to get rid of them.


- Michael

Hi Michael,
Thank you very much. I guess I didnt look at it very closely, I thought I looked for that kind of stuff. But then again, 0.0000007 is very small and perhaps I didnt look close enough!!! LOL

Not to drag this out, but just a finish up for me... It appeared this iges file came from NX also, and I was wondering about this poor geometry creation. It appears to be a simple hole punched with a fillet on the edge. Could this type of issue be getting created by the filleter mechanism? I mean, when I punch simple holes like this in a revolve with MoI and fillet the edges, When the fillet succeeds, I get good surface quality. I realize that if I create a poor curve to punch my hole, this would create this kind of issue....But if I create a simple circle, it would surely be avoided!!! Could it be a loose tolerance setting in NX that would allow it to create these anomolies by just "slapping things together"?

I know there are many variables, and you are not a mind/model reader...Just looking for a "best guess" situation or more brain picking on your judgement of what has happened here....Perhaps these are just bugs in NX5 as that is old now compared to 7.5.

Thanks.

Hi Burr, well it's hard to say what's created that.

Maybe it's even something that the IGES importer itself is inserting in there because it thinks that there is too large of a gap in there, I'm going to see if that is a possibility.

- Michael

Hi Anis,
Well I got to look at the model and received the same results as everybody else. One thing that the higher end packages do, with these repair wizards is list all the faults and present them to you to fix. With the package I used it would present about 50 areas that need to be addressed. Perhaps a day of working the model over by someone experienced.

Anyway, thanks for letting me look.

Is this for rendering purposes?

No, I want to create core & cavity.

Hi Burr, so actually that particular line is being created by the IGES import code.

The reason why though is that the file is marked as being built to a tolerance of 3e-7 (0.0000003) and the IGES reader seems to take it at its word and considers gaps much larger than that to be large enough sized things to be filled in, while in actuality the data in there is nowhere close to that accurate, more like 1e-3 between endpoints and stuff like that.

This tolerance value is at the start of the file here:



If you edit that to say something like 3.037E-003 instead, then at least some of those micro lines will be avoided from being created, but there still seems to be a few other ones in there apart from that.

I'll have to send this over to the HarmonyWare guy and ask him if its a good idea to actually trust that tolerance value in the header since it is completely off in this case.

- Michael

Hi Burr...

Thanks for looking at the model to examine.....
So the conclusion is I need to repair some bad area manually or request to our client STEP format.
Also looks like no software that can import the IGES file smooth and solid.

Thanks for everybody that help me to solve this case...

Regards,

Anis