Finding Equation of Imported curve

[scorpion15]

i've a problem regarding finding the equation of a specific curve of an imported feature. i want to use that curve with some extension on both sides following the same equation of law which the curve is already following but don't know how to do that. any ideaz please!

 

[kenppy]

do you have coordinates for the curve or is it just a 'dumb' import?

 

[s_pme20]

could you post image showing shape of curve?

 

[scorpion15]

its a 2D curve and i get it from referencing the outer periphery of the part!

 

[kenppy]

if you copy and paste the curve it gives you the option to extend, this works pretty well as long as the extension is not too much as a proportion of the original.

 

[scorpion15]

isn't there any way of finding the equation of that curve?

 

[kenppy]

post the part or surface and let us have a look at it.

 

[tobbo]

whats the goal of adding extra length to your curve? And more specific, why do you want/need the equation?


I would also suggest to do as kenppy describes. Copy/paste - draghandles.


//Tobias
Edited by: tobbo

 

[scorpion15]

i am making a program for ultra surface finish on 3-axis lathe. the tool path needs to be defined by that curve extending the both edges of curve which i am using as reference. because the required surface finish needs to be in nano meters, so i can't just drag the ends. i need proper extension following the same rule as described by the part reference curve itself!

 

[kenppy]

how will you know if there is a unique solution to this? do you know for sure that there is an equation to this and not free-from?

 

[scorpion15]

unfortunately i am not sure about that whether its a free curve or there is an equation behind it

 

[kenppy]

i'd still like to see the curve if you are at liberty to post it.

 

[tobbo]

Interesting, i assume that you want the toolpath for the mill to travel "outside/past" the curve and so on? Maybe back onto the curve again, and thats why you want the extension of your curve to have the same shape as the original curve?


If thats the case, you would be fine with a tangent extension (or curvature extension if you would prefer that)


But maybe i didnt understand you correctly?


A picture would be great.


//Tobias

 

[Bob_W]

If you can pull some X-Y coordinates from the curve, there may be a slim chance a curve fit program can come up with a simple solution using those points as input.

Bob

 

[cpoirier]

$0.02 offered here.
http://easycalculation.com/statistics/learn-regression.php
http://www.curveexpert.net/
Regression or some software to do it for you may help.
Every curve has an equation..... your using a computer.

I have a crazy idea, make the curve (project a line onto the surf), export to IGES or STEP, get the standard for the translation method so you can hack the file, e.g. sift through the text and find the curve and the parameters that make it, I assume the standard will have the equation to rebuild it. Like I said crazy. DXF standard is easily found on the Autodesk website if you look. Using an x-sec may or may not do it for you. I used this once to take info from a CAD design and use it in some home grown software. (Perkin Elmer)

Now having said that, once you are off the surface it is not be likely to be correct, the x-sec is probably approximate, likewise you may loose fidelity when trying the export hack. In the end.... a few nm! .000001mm??? Are you using a laser interferometer on the machine??? If your doing this you are already crazy so cracking a STEP or IGES is not insane.. just crazy. If it came from an optical modeler you already lost something in a translation as it was an ACIS, IGES or STEP to begin with. (zemax, Trace Pro, etc)

Sounds fun,

Chris

 

[scorpion15]

thanks cpoirier, ur idea is really crazy, but i already made the programe yesterday using the tangent extension. the part is now finised and the mirror shine is great!

.

 

[zpaolo]

cpoirier you idea is crazy but should work, on a theoretical point of view. What you'll obtain is probably a NURBS curve or a spline curve, that is not exactly an explicit curve but still can be treated mathematically with a proper framework. That said, Pro|E does usually a bad job when exporting iges, because he downgrades all NURBS curves to degree 3 curves, even if you just open an IGES with curves of higher degree and save it back, pro|E will lower the degree and add spans and approximations to the curve. All of this without considering that, as you said, usually you are already approximating when a certain mathematical surface is imported in pro|E.
Fitting a curve points with a formula is a very complex, risky and error-prone task, if we are talking about optical parts, then IGES files and CAD models are seldom used, usually it works the other way: you transfer the formula details to the manufacturer.

Paolo

 

[cpoirier]

Paolo,
Do you have the IGES and STEP specifications or know where to get them?
Sounds like you know some of the inner workings better than most.

I often wonder if there is a way to write a configuration file to tweak the export. Have you noticed that when you export from Pro there is an "options" button and if you click it it looks for an config option file? Another mystery from PTC. I wonder what non-published or supported options exist. A litle mystery is fun unless you have a deadline.


Thanks for your comments.
Chris

 

[zpaolo]

Paolo,
Do you have the IGES and STEP specifications or know where to get them?
Sounds like you know some of the inner workings better than most.

There is a public specification for IGES, you can find the link from Wikipedia, and it's http://www.uspro.org/documents/IGES5-3_forDownload.pdf The same is true for STEP i think but IGES was simpler and I never looked for the STEP specification (STEP is more of a "product" descriptor while IGES concentrates on the geometry). Actually I had to study the specification thoroughly (you'll be surprised to read how many parametric and parent/child aspects there are) but never found a way to tweak the export from Pro|E. I used what I learned to export revolution surfaces directly from Mathcad to an IGES file, and to import them in Zemax for raytracing, therefore I know more about curves (from conics to NURBS) than about surfaces.

Paolo

 

[telecomguy]

I've used Excel to provide an approximate curve fit before. Align some points to your curve and extract X,Y data. Put that data into a spreadsheet and add an XY plot. Right-click the curve and you can 'Add Trendline'. You can try the different options such as Nth order polynomial. Then click the checkbox to 'Display Equation on chart'. Depending on your curve, you may get a good fit.


View attachment 5483
Edited by: telecomguy