Sweep and Twist in Unigraphics

[iucu]

I HOPE sameone here could help me. I have to design a stator winding. Consider a BAR (section 6mm x 1mm). I have to bend this bar, I mean to Sweep (along a guide) and Twist (normal to guide). I was using Swept and Mash Surface with bad results (section I got is not constant along the guide and at inflexions points ugly shapes).

 

[iucu]

Again, the 2D profile is moving along the path and in the same time rotate around the tangent vector of the path. I have UG NX5. Is like going up to hill, to peek, in one side and down in the other side, so at peek is a twist.

 

[JohnRBaker]

Here's an example of something that might help you get an idea of how to approach your problem.

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]

http://www.eng-tips.com/viewthread.cfm?qid=216683&page=5

We all had a go at this one last week

Cheers

Hudson

 

[iucu]

I can not open the file. My mistake, I have NX3 not NX5. Could you help ?

 

[iucu]

This is what I am trying to do, see picture

 

[hudson888]

This looks like some sort of flat wiring harness package or similar conduit. I'm wondering how accurate it needs to be. One of the problems you have with sweeping long and complex continuous section is that unless you've an extraordinary degree of control you can run into problems with the section orientation relative to the guide curves and have the section effectively change its shape and size. This is in fact a feature that is used deliberately to describe sections with taper or enlarge or sweeps through more than one changing sections.

In your case the parameters that you specified seem to have not controlled the sectional orientation of the feature.

Also you have long straight or near straight sections turning sharp corners which are too sharp. So if the approximately vertical and horizontal sections meet at right angles then you can have as tight a radius as the material can be engineered to accept. But if the corner meets a a kind of mitered angle then either the radius needs to increase or the section CANNOT remain constant it either changes or is expresses some local distortion or compression to represent the elastic behavior of the material. Now if and when you do the later and expect to return the section to its former profile after the bend then YOU have to control it. As you can seen when uncontrolled the whole shape flattens. To do that I might build the model from a series of sweeps or I would make some changes to the path to avoid tight corners and yes I would work on ways to deal with the twist in the middle, perhaps using techniques that we worked with in the examples last week.

Give us a model and I may have a crack over the weekend.

Cheers

Hudson

 

[JohnRBaker]

OK, take a look at this model and see if it will work. This is about as simple as you can make it, in NX 3. The good news is that the solid model itself is very simple and well-behaved with smooth faces and edges.

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]

John,

I have attached my own NX-3 version of your model after a did some experimenting with it. Firstly I should say that your method is almost exactly the same as mine. I posted earlier about the image we were left to work with, and what I thought to be some things that showed up as problems with the construction, and I started work on something similar breaking it up into several sweeps.

The image appears to have radii on the four edges at the corners of what is basically a rectangular section. As inferred by your model I too would have applied them after the sweep operation, and therein lies my only problem. Your model did not easily take the blends, because the center sweep lacks for good tangency where it meets the other two. I have used a little trick that always works for me in NX-3 I created a short line prior to either end of the bridge curve so that the sweep is better squared up to the end of the other two. You can hardly pick the difference but the tangency is very good indeed.

Now I have a question. I loaded the same model yours and mine into NX-5 thinking that it would be easier to analyze the sweep construction method using the newer dialog. What I noticed right away was that when you edit the parameters of an NX-3 created sweep you get the text based menus, but when you edit those of newly created sweeps you get the new more graphical version. And the sweeps are different geometrically. Only slightly but they're not the same. In fact they're worse with regard to the tangency problem so that even if I increase the length of the lines starting either end of the guide bridge it doesn't do as good a job. The starting end in particular seems the worse in this case.

Could you please look into this using the examples and see whether I'm either doing something wrong or something else can be done?

Best Regards

Hudson

 

[JohnRBaker]

The menu issue is that the new (starting in NX 4 I think) swept feature is different than the old one and so we have to leave the 'old code' in the system to allow you to continue to edit any 'old' swept features. That's what you have to do when you can't perform a 100% conversion from one version of a feature to what one would expect the newer feature to be. This is not as rare a situation as you may think and it's the only alternative to making old features 'dumb' and treating them as just bodies.

As for the differences in the model if you were to start over, you may just have to take a different approach altogether. After all, your NX 3 'solution' was not perfect either, was it?

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]

John,

Your last sentence is on my mind. After I united the three solids I ran examine geometry to 0.02 and 0.1 for the angular, which is much tighter that I would normally use. The model passed every geometry check except tolerances with tangency errors highlighting only those edges were at this stage left sharp. This was in NX-3, and at this stage I'm thinking that perfect is too big a word but it meets the criteria for tangency. But wait there's more....

I then applied a blend to the edges all four corner in one go, ran examine geometry and it game me Face-Face Intersections and Self Intersection errors. But then after I ran the blend as two blends two edges in each it the right sequence I managed to get a result which although apparently geometrically identical does not have those geometry errors. Both blends were created with remove self intersections turned on. A redo of the blends with remove self intersections turned off made no difference. But wait there's more....just a little...

So a saved the model with the blends applied to all four edges being the case that failed geometry checks in NX-3 and re-opened that in NX-5 to prove that it still failed geometry checks. It did, but all I had to do was delete and re-apply the blend to all four corners and regardless of whether remove self intersection is checked in the blend dialog the model now passed examine geometry.

Now I'm thinking that the only place that I'm seeing potential for problems has to do with the tolerance errors where the three united sweeps are joined together. I've never been 100% sure if tolerance errors anyway, how to get rid of them often eludes me. In this case I haven't seen a model do these things before, but I will say that regardless of the twist, I find long sweeps with multi segmented guide curves have always been a problem. We have difficulty getting the section to accurately hold its dimensions and orientation as the number of segments increase. I imagine that somehow the tolerances affect the outcome, but I have found nothing that I could set to improve matters. However if this means that we have to split up the sweep into do-able segments then we need to maintain tangency across the boundaries between faces, otherwise we do have to find other methods and I do find that disappointing, and even more so when I find NX-5 seems to have taken a step backward from NX-3 in that regard.

John I don't necessarily expect you to comment on my observations in detail I'm sure you'll work on the code in the background to keep improving future releases. But since you hinted that my model was less than perfect I don want to get a better understanding on what you meant by that as regards the observations I was able to make about the blends etc whether those were what you had in mind, or whether you had other concerns with my method.

Oh and what are we supposed to tell the poor guy who posted with the problem in the first place? I think we had better recommend some other solution. Perhaps the twist and sweep in the dipstick method may work.

Best Regards

Hudson

 

[JohnRBaker]

Perhaps I used the wrong term. I'm sorry if that offended you as that was never my intention. What I was referring to was that your model needed some tweaking to make it work, as would have mine if I had gone to the next step. The point was that everything is done with tolerances in mind and we have to accept the fact that it's OK to stop before it's perfect, not because perfection is not worth achieving, but because the world does not require perfection, it only requires that it be good enough. The trick is to know what's good enough and when you've achieved it.

With that in mind, since we did not have any actual specifications, only 2 parameters and a picture, all I was trying to do was get as elegant a solution as possible since I knew it was going to require a procedure and I wanted as few steps as possible. Besides, what was being modeled would never be manufactured from the models, but rather fabricated in a process that will bend and form copper buss-bars to form the 'winding' of an armature, which BTW I've seen something like this before at a customer site in Texas a few years back. At best some tooling, a forming mandrel or anvil, will be created where copper buss-bars will be basically hammered into shape. This is more of 'blacksmith' job than anything.

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]

Thanks John,

Did I get over defensive, sorry not my real intention, enough said about it that I intended to defend my model only insofar as to direct your attention to something I imagined you had guessed at but not discussed. Because the sentence ended with a question mark so I guessed there was something more to be found if I looked for it and indeed I did find a couple of things that frankly surprised me a bit. I really wanted to talk about the geometry errors and the tangency think looking to learn how to make better sweeps.

Yes you could probably see from my earlier post that I compared it with other similar types of things that rarely need to be particularly accurate. Since the process takes over at some point there will be some uncontrolled deformation etc., just not as uncontrolled as the image we were given I think you would agree.

Anyway I guess you'd answered my questions as far as it goes and if you've any other hints about bettering the art of sweeping sections I'm sure to remain interested.

Cheers

Hudson

 

[iucu]

Thank you for all responses. So far it is a improve but I still have to do more.In pictures you can see how was before (right side) and how I got it so far. Thank you again