Cancel
Showing results for 
Search instead for 
Did you mean: 

Re: Spline approximation of wire import

Creator
Creator

I guess so. The geometry is in fact the result of a 3D scan consisting of polygonized points of which I have taken a section and exported it as IGES. The cross section consists of 585 discrete points and I guess they have just been exported as points then?  Edit: I checked the IGES file and it did indeed just export 585 discrete points. Is there a way to connect these points or should I try to approximate these points through splines?

Re: Spline approximation of wire import

Gears Esteemed Contributor Gears Esteemed Contributor
Gears Esteemed Contributor

I'm still not clear. Your last attachment showed more than points.

Bruce Shand
SE2019 MP8 - Insight - Win10 - K4200

Re: Spline approximation of wire import

Creator
Creator

When I zoom in on the part geometry, the points seem to be connected, but when I want to create a revolution, apparently they aren't? This image is clearer hopefully.

Re: Spline approximation of wire import

Gears Honored Contributor Gears Honored Contributor
Gears Honored Contributor

Don't remember if V20 allowed revolving connected segments without relationships. The sketch in the image does not seem to have relationships, though latest versions as far ST6 allow revolving a series of lines with connected endpoints having no relations.

  

~Tushar

Re: Spline approximation of wire import

Creator
Creator

Is there anyway I can approximate this series of lines with a spline in V20 and revolve that?

Re: Spline approximation of wire import

Gears Esteemed Contributor Gears Esteemed Contributor
Gears Esteemed Contributor

Hi there Hendrik,

 

I don't know of any automated way, in Solid Edge [I can do this in another program though]....but I suspect in SE you'd need to create a new sketch and use the end points to create new geometry, being careful not to use too many of the control points. [either by tangent lines & arcs, or a continuous spline curve]

 

Also, feel free to upload your ".PAR" file for myself or one of the good people here, to offer more exacting help, based on how the geometry will respond and get to the result you need.

Sean Cresswell
Design Manager Streetscape Ltd
Solid Edge 2019 [MP8] Classic [x3 Seats - Cloud Enabled]
Windows 10 - Quadro P2000

Re: Spline approximation of wire import

PLM World Member Legend PLM World Member Legend
PLM World Member Legend
One way is to simply approximate the imported geometry by drawing a new spline on top of it but that will give abitrary accuracy.
If you actually have to get the curve to pass through all the points it may be possible using some other package to clean it up and then re-import the curve but this would be computationally expensive.
A third option might be to approximate the points using ST8. The tools to do this are much better than in V20. If you were able to share the point file I'm sure one of us could do the curve and export it in Parasolid or STEP format suitable for import to V20. You'd need to let us know roughly how accurate you need it to be.

Frank
Edinburgh University

Re: Spline approximation of wire import

PLM World Member Legend PLM World Member Legend
PLM World Member Legend
Sean has said exactly what I was thinking! My last post was put in before I could read Sean's!
Edinburgh University

Re: Spline approximation of wire import

Creator
Creator

Many thanks for your interest and effort!

Attached I have included the draft file, the part file, and the IGES export file containing just the points. The accuracy of these points is highly important, the measurements happened with a relative accuracy of 0,01 mm. Also the distance to the Z axis is of great importance. I apologize for asking such basic things, I am a PhD student in welding engineering with little experience in CAD software.

Re: Spline approximation of wire import

PLM World Member Legend PLM World Member Legend
PLM World Member Legend

Hendrik,

as is often the case, what seems like a simple question is actually quite complex.  What method you use to fit a curve to this data is likely to be dependant on what you are going to use the curve for.  Apologies if you already know all this but the question of accuracy is quite difficult to define and answer.

You could fit a curve through all 500+ points but it will be very slow to manipulate because the curve would be so comlex and time-consuming to compute. Given that there are already small errors in the point definitions if you forced a curve through all points it could actually be very innaccurate.  The curve could be within the 0.01 band of the 'real' location of the points but between points you have no idea.  The curve could end up wavy and thus a possibly poor representation of the actual 'real' shape.

If you were to use the resulting curve to calculate the surface area of the shape it might not be within a known tolerance. Similarly if you used it to be part of a CFD simulation the waviness could badly affect your results.

Most of the time CAD engineers want a smooth curve but it is difficult to mathematically explain how 'smoothness' is relevant to an analysis package (CFD, FE or whatever).

I hope all this makes sense but tomorrow if I get a chance I will do a few tests on my work machine to try out a few different curve fitting regimes.

 

Frank

Edinburgh University