Equation-based curves in Sketcher

[treddie]

<DIV =wordwrap>Hello.

Is there any way to build equation-based curves in Sketcher.I know how to do this when building a DATUM curve, but it seems odd that ProE does not offer the same functionality in Sketcher (say, when building a sketch from within a Variable Section Sweep).The datum curve approach gives you the option of building the curve by clicking "From Equation".Sketcher does not.And building a relation does not offer the ability to "sweep" out a curve based on the parameter "t" as with the datum curve's "From Equation" option.

The reason I need this is that I have a closed curve that much change shape (or size) according to the value of trajpar when building a Variable Section Sweep.But building a datum curve first doesn't cut it because the VSS can't access the contents of the datum curve to alter it via trajpar.You can only use Sketch > Edge > Use, which means you're stuck with the dimensions of the datum curve as previously built.

The only other option I can think of is to use secondary trajectories for the VSS to follow, but that opens up a whole can of worms, and in my case, a zillion extra trajectories to accurately shape the feature I need.It also means I would have to approximate the original curve by a spline, and this then makes the construction inaccurate and prone to interpolation errors.

Thank you for any responses,
treddie </DIV>

 

[jeff4136]

Are you up for exloring options?

 

[treddie]

Hey Jeff! Nice to hear from you, over here.


I'm open to anything now.


My latest attempt just now was to build an IBL points file in VB6 (which is actually where my test curves originated anyway), import that into ProE, export out as IGES, and reimport that into Sketcher during the VSS. That gave me a scaleable curve that reponded to trajpar, but I hadn't gotten to the point yet of getting its parameters to scale properly yet, due to the params ProE selected on its own when the IGES was imported.


But I'll drain my head, for open input, and go with the flow.


treddie

 

[jeff4136]

A few ideas bouncing around, none really pretty at first blush.Which might be good candidates would depend on the curve (shape ... airfoil?) and how it changes (not a linear taper or twist; something a linear blend would be appropriate for?).


Fitting sketchercurves over the shape may be the only option. I'm up for some experimentation if it comes to that.


What can you tell or show us about it?

 

[treddie]

Basically I have an elliptic cone that tapers from a given size down to a point. But it is not REALLY an elliptical crossection...it is an ellipse with a filter applied to it, so that the equation has changed drastically. The shape hasa subtle squared off feel to it. I wish I could share the equation, but it is part of a program I am building for eventual sale, and the equation is a bit proprietary. But the shape cannot be approximated...it needs to be what it is based on that equation. Please see attached Photoshop approximation, 2008-04-17_195804_Boxy_Cone.zip.


In principal though, any equation based curve will do. I just don't understand why ProE is so un-user freindly when it comes to generating a solid based ona scaled and propogated surface of ANY crossection. Seems like that would be one of the FIRST things a solid modeller would support. After all, it is just a matter of scaling a surface at points along a trajectory. Awfully simple.


So think of ANY shape of surface(imported or otherwise) that is made to change scale along an origin trajectory. In this case, it just happens to be a cone.

 

[treddie]

I forgot to mention that in my earlier post, I was referring to a zillion extra trajectories. That would certainly be the case if I were doingthe VSSwith a very complex crossection and using controlling curves to affect its size along the sweep trajectory. I was thinking in general for ANY crossection, not just the simple cone I am working with.

 

[treddie]

Jeff,


I THINK I have the solution, and if I'm right, it's a lot easier than I suspected. Not as easy as 1,2,3, but definitely not a headache. I went back to your ellipsoid helmet file and took a closer look at the curves you used for the ellipsoid sector you built with a VSS, and it dawned on me that you only needed to use 2 control curves and that having done that, you still had true elliptic crossectionsthat were correct from one edge of the sweep to the other. I'll write back when I confirm that I do indeed have a consistent sweep going on. But it still involves the IBL/IGES setup before I start the VSS.


Nonetheless, If you have any new takes on the problem, I'm very interested in what you drum up, if you feel like taking a crack at it and you have time to spare. Especially if it ends up being a more efficient process.

 

[jeff4136]

The only one-size-fits-all that comes to mind is to continue in the same vein; generate points for the entire surface and create the surface from an .ibl, assuming there isn't some limitation to prevent it.


But ---
Your pic appears to be corrupt, however for the given description (scaled section, cone) a Boundary Blend between a curve and Datum Point should be appropriate.


Just to clarify; when I speak of fitting sketched curves I'm talking about attempting to fit to near system tolerances or, worst case, well within normal(?) manufacturing tolerances. Difficulty is rate dependant, some curves aren't especially difficult.


And an additional thought; if you can define the curve with a time function parametric equation a native curve by equation will probably be preferable to an interpolated curve thru points. The system will only generate enough knots (points, not equally spaced) to fit the curve to system tolerance.

 

[treddie]

I didn't think about a boundary blend to a point. I'm going tp experiment with that.


"if you can define the curve with a time function parametric equation a native curve by equation will probably be preferable to an interpolated curve thru points. The system will only generate enough knots (points, not equally spaced) to fit the curve to system tolerance."


I definitely agree, especially after the helmet debacle. But I could not find a way to build a time dependent function within Sketcher (which still amazes me about ProE). But going with a boundary blend to a point is about as simple as it gets. Now if it ISN'T a cone at all but a more complex sweep, I can see that your VSS approach with 2 equation-based control curves (again, your helmet file), is appropriate. The only caveat being that the section ends up being based on points connected by inefficient (and prone to tolerance errors) splines, rather than efficient splines coming forth from an equation.


But now that I'm thinking about it, now I think I see what you're saying about the Boundary Blend...I can still build my equation-based curve as a datum curve and let the BBlend handle all the sections, rather than resorting to a VSS which relies on a sketched section. There was a guy over at ProE Eng-Tips who was pressing for a BBlend, and now I see that I wasn't understanding (once again) the power of a BBlend.


So much to play with...so little time.

 

[treddie]

Yep, that worked. The BBlend I mean. I built the equation curve and the only problem I had (which had me scratching my head) was an issue of "boundary curves not tangent to tangent surface". Once I figured out what the hell ProE was talking about exactly, that was a quick fix by turning the equation based curve into a spline sketch (Sketch>Edge>Use, and Copy/Paste to break the alignment with the underlying equation based curve), thengetting the problem end of it tangent to a centerline via constraints(and therefore perpendicular to the plane normal to the boundary on that side). Floating point errors were enough for that end to register as non-tangent. Nothin much to look at here, but here it is...2008-04-20_013125_Boxy_Cone_2.zip.


Once again, thank you for all your help, Jeff.