Working with surfaces 1

[ddf0123]

I recently got a parasolid from another program. As far as I know, it comes as surfaces.
When the file came, it was composed of 5 different sectors (spilt with an edge) but one unparameterized feature.
My client wants to merge it all so that no edge will show and then I need to edit it slightly. Project a few curves here and there.
In my line of work, I have never encountered a need to use surfaces.

Any suggestions as to how to edit an unparameterized surface?

 

[Xwheelguy]

Some posting tips:

Always always always always always always always post the version of NX you're using. Like most softwares, different versions have different commands available to the user, and that will affect any responses you might get in regards to your issue.

Please please please please please try to use terms common to NX. Sectors, split, merge and edit it slightly mean nothing, to me at least, in terms of NX commands. Please be more specific and describe how you plan to change or edit the model. If you can attach the NX file or an image, that would help tremendously. The more we know, the better we can steer you in the right direction.

Tim Flater
Senior Designer
Enkei America, Inc.

http://www.enkei.com

Some people are like slinkies....they don't really have a purpose, but they still bring a smile to your face when you push them down the stairs.

 

[jerry1423]

You can try to sew them together and create a solid from that, but it would be nice to see what you are working on.

Is it possible to post a shaded image of what you have?

 

[hudson888]

The question reads as if the unparameterized feature might be a sheet body and not a solid body. That is to say if there is only one feature then you could not envision the need to join anything. Presumably you have a sheet composed of several faces which is either equivalent to having already sewn them together or could be as a result of another CAD program which produced multi patch surfaces that get rendered as separate faces to NX by the translator.

If this is the case then you could try un-trimming copies of the sheets. Hint they may all be identical with seperate boundaries. This is somewhat unlikely, but quick to check.

For the same case where you perhaps have to rebuild several sheets into one. The suggestion that you make as it seems to me is that you need to rebuild the surfaces. Firstly perhaps we should start with asking why, and a look at your geometry would be invaluable for that.

Otherwise when it comes to surface rebuilding tools you could try a through curves surface, or using curve mesh, even the studio nxn all of which use curves and edges that you'd take off your existing geometry to rebuild something approximately but not exactly the same over the top of the imported data.

There is another tool called a quilt surface which is intended for similar such tasks. You could try it but I'd warn you that in my experience it seldom gives great results.

Best Regards and sent us something to look at

Hudson

 

[JohnRBaker]

Of course the question has to be asked, why is it so important that these 'internal edges' be removed and that you have only a single face? Is there some operation that you wish to perform that will not allow you to select multiple faces or is it just some aesthetics issue of not liking the idea of having a subdivided face?

John R. Baker, P.E.
Product 'Evangelist'
NX Design
Siemens PLM Software Inc.
Cypress, CA

http://www.siemens.com/plm

http://www.plmworld.org/museum/

 

[hudson888]

Assuming my reading of the question to be correct. As John says, I can and have posed exactly the same question. It is a very good place to start.

How to ask that question, is to ask the how and where your client wants to look at it in such a way as to NOT see internal edges. After all you could just not show edges in you shading settings in model space. They never show up unless there is a discontinuity in the surface when you create high quality images. In drafting you can simply turn of smooth edges in the views.

Cheers again

Hudson

 

[ddf0123]

Ok. There are some issues that need to be answered.
Even though at home I have NX5, the version I am using with the model is NX2 (yes, it still exists).

First of all, it will not be possible (as much as I would prefer otherwise) to post any sort of picture. I am using UG in a military base and we aren't allowed to take things out. If it is essential, I might try to bend a few rules.
Secondly, the model that I am working on (which isn't part of what I do on a regular basis) is an exact replica of a Tibia bone. It was imported from another program. Not a product of UG.
The model has to be free of any edges because my client is going to run the model throught an analitical program and the edges will complicate it further.

I would like to hear some more comments. Hudson's first comment has some excellent tips.

 

[JohnRBaker]

You may need to extract a mesh of points from over the area of the model that you wish to convert to a single face and then use reverse-engineering approaches to create a new single surface using the points as input, such as Surface from a Cloud of Points or some such thing.

John R. Baker, P.E.
Product 'Evangelist'
NX Design
Siemens PLM Software Inc.
Cypress, CA

http://www.siemens.com/plm

http://www.plmworld.org/museum/

 

[hudson888]

My post went astray,
I had explained the following at length.

Firstly you can probably safely assume that the bone is intended to represent an average so that you have some latitude to rebuild to a tolerance, and therefore simply rebuilding the surfaces with the most suitable tool would be the appropriate thing to do.

I wonder about the analysis whether you need to inquire further in order to better address some issues I mention below. In my experience FEA at least is capable of meshing segmented models perfectly well. They could be putting you to more trouble than it is worth.

Based on the nature of what you described there is still a small unanswered element. You see if you want to create as near to an edgeless surface as possible then maybe you could imagine something like an open ended tube of varying section describes it fairly well. The problem with that is that it is a two edged surface and most surfaces need at least four edges.

Both the quilt as I describe earlier, and the cloud of points are methods that rely to some extent on projecting the points onto four sided boundaries or other surfaces. In part because what you're dealing with creates surface normals that more or less radiate from a central axis I doubt that either method will succeed. You welcome to try, but I think that the projection method in the quilt for example is prone to mapping the projected normals right through to the opposite side of anything that wraps more that 180 degrees. In fact 90 degrees is a much safer place to limit it.

The probable best available method for a two sided surface is a through curves surface using radial sections, or rings if you like. These could use existing edges from your original model or you may choose either to cut sections or extract isoparametric curves off the existing surfaces. Having accessed some initial curves you may do well to rebuild or edit them so that the meet end to end with at least tangency, (G1 continuity), within each section. It will also serve you well if they are closed curves or chain-able with end points that roughly align from each section to the next. Building the surface is simply a matter of selecting sections so that the arrows all point in the same direction. If you have enough sections you can increase the degree making the surface more complex and possibly following the original more closely.

You may get an nxn studio surface to work with one longitudinal guide, but if you lack the license then don't worry as it will probably not be a great deal different than the through curves method.

If you're willing to permit longitudinal seams then a four sided surface such as a curve mesh or one of the nxn studio surfaces would be most suitable for rebuilding your model if all of the above fails to work. with either of these you can force tangency along the seams.

Best regards

Hudson

 

[basenxuser]

i too have had the thought for a combine face option, which will help in 5-axis CAM.

Adding my cents to the discussion..If you are able to find a nominal surface normal for all the splits, then try using QUILT option

HariharanB

 

[hudson888]

ddf0123,

The attached is a tibia model obtained from the site below, secret what secret?

http://www.3dcontentcentral.com/default.aspx

Now I have remodeled it in NX-5, I could go back to V18 but I'd have to be bothered and in this case you'll see there's no real need. I have also included the original Parasolid, and another suitable for use with NX-2.

I gather you'll be able to open the thing in NX-5 and at least have a look at how it is built.

The curves that I created were isoparametric, and I used join curves on each section and then created a through curves surface. You turn off preserve shape in NX-5 to lose the seams, and there it is. QED.

Cheers

Hudson

 

[ddf0123]

I managed to get the parasolid out of the army and I am working on it at home, where i have more power. I tried running the quilt command and the computer just crashed.
That means that I can now post a pic.

This is the first time I am posting a picture. A few times and I might get it right.
Here you go...

 

[ddf0123]

Didn't get throught. Do I have to compress it using winrar and send it through there?
By the way, thank you Hudson, once again, for a superb attachment.

 

[ddf0123]

As I stated before, surfaces is a whole new subject for in UG and I am having a little bit of difficulty.
I decided, to keep the body as accurate as possible, that I will place a datum every 5 mm. I then extracted the curve for every datum. I tried using: insert->mesh surface->through curves, and: insert->mesh surface->studio surface. Both gave me unsatisfactory results. The surface was twisted beyond recognition.

Any further advice?

 

[hudson888]

The attachment never came through. The need for using winrar relates mainly to file size, wherein if the file is over 2Mb these kinds of uploads seem to suffer. You can attempt to download your own upload after you create a post, so that by testing it yourself you'll be sure that it works. When posting between time zones on web forums like this one you can save a day that way.

Are you doing something basically similar to my posted example?

Are you attempting to get rid of, longitudinal or radial edges from the original?

For this kind of surfacing work most people don't get hung up on associativity in the creation of the surface. So we treat the surface as a base feature a lot of the time and don't always keep any parameters. There are exceptions but they are relatively few. So I wouldn't bother with datum planes if I were you. In my example I extracted isoparametric curves simply imputing that I wanted 50 in the V direction. I did so separately on either side and almost miraculously the ends of the curves on either half happened to line up exactly. It was no fluke really since it happens often that if the surfaces are built the same way on either side then they will divide up very evenly. In your case in the original data is less regular then just take a series of sections at even intervals which you can do in one command to extract some curves which you should be able to join to form closed sections during the same operation if you wish.

When it comes to accuracy in my example I was able to get within 0.083 mm of the original. Which for most analysis purposes should be more that close enough since the meshing process performs some approximation in any case.

In the case of what you're doing my only concern is that your bone may take in the joint on either end which could present some other real difficulties. I'll leave you to think about that because it may not be necessary to worry about at this stage.

Oh by the way curve mesh will only do four side surfaces. You would have to accept seams per the data I downloaded at the beginning of the example that I sent you. If you do persist pick the curves in order and keep all the arrows pointed in the same direction.

Best Regards

Hudson

 

[ddf0123]

I decided to best let you see the probelm to post the parasolid

 

[hudson888]

So how many surfaces do you propose to have in your final result. My earlier result had three, and your current model has six. Two are easy to get rid of using my earlier method, but the ends are reasonably difficult to address. That is partly because of the shape and also because the original lacks tangency.

My earlier example was prepared in under 10 minutes, simply because I elected a method that worked very easily for that geometry. Of course you charge an hour for the job, 10 minutes for doing the work and 50 for knowing how! .

Let me know what your needs are and I may be able to assist further.

 

[hudson888]

It came to me in the meantime that three surfaces is about the least that I think you can use. The attached is reasonably accurate, and the technique similar to as used for the other example. The top and bottom patches of surfaces are just a couple of cloud of point surfaces, as this surface type has no parameters you could not otherwise have known that from my models. Nor did I bother to keep the points for file size reasons mainly.

I didn't chase any additional surface refinement or even great tangency as I doubt there is call for it. Normally when we talk about surfacing well we mean A-class, which is a completely different discipline where the criteria are almost diametrically opposed to what you asked for,(i.e. you're able to create as few or many surfaces as you need to describe the best looking shape). I just didn't want you to think this represents what I'd call great surfacing practice in most disciplines.

Cheers Again

Hudson

P.S. If I'm going to do the whole job for people I'll have to start soliciting donations to my favorite charity or something

 

[ddf0123]

Let me just state that you are a god in my eyes.

If this is the help you normally give others on the forum, then you must be really well liked through-out. I am sure your work place must love you. In Hebrew, using teenage slang, someone might even say Ata gadol!

Your first method really was most accurate and I want to try sloving the problem using that. The bone forms well, until after the peak. Then the sheet begins to bend into itself. Is there a way around that? I tried creating more curves to make the transaction smoother, but nothing I do seems to have any effect.
So for the last time....help?

 

[hudson888]

I don't want to go anywhere near competing with the God of the Hebrews. Lets just say I must be an atheist since I don't believe that of myself. Anyway you're probably overestimating since I haven't spent more than a hour at a time on your job just for my own interest.

The second method is very similar construction as the first one. I think that you can't get below at least three surfaces. And I think that you have good enough accuracy. To get greater smoothness smoothing the curves can help, but certainly using less curves rather than more is the way to allow more smoothness. I say allow because you generally allow smoothness by sacrificing accuracy.

Now you possibly need to comment more on the second method as it follows what I believe to be your actual data. If you clearly state the number of surfaces and the tolerances that you require then we can talk about the possibilities.

One thing to consider is that you don't want to use other than the through curves method if you need to avoid seams in the longitudinal direction. Another is that the capping surfaces of either end could be rebuilt but not as for the main surface derived from isoparametric curves. The problem with the isoparametric curves around the top and bottom of the shape is that they describe circular topography (like a map), meaning that there aren't four corners. You would want to section the ends to get a four cornered surface in order to build it as a curve mesh. Even then your are correct in observing that you can't get great tangency to the main surface that I built, which in large part is due to the fact that it isn't smooth enough.

I sense that you probably want to understand the way it is done and achieve this task for yourself. If you need more help then post again, but be quick and certain about what you're after.

Cheers

Hudson