V5 equivalent to V4 arc 1

[jackk]

Is there a way in CATIA V5 to make a smooth curve that's a best fit through a series of points?

A spline curve is too bumpy and passes exactly through all the points.

 

[solid7]

Well, a spline is actually the definition of a "smooth curve that's best fit through a series of points."

I guess the key word here, is "through." If you want to take a series of points, and "best fit" a spline, that's a different story.

If I remember correctly, there were 3 spline optimization options in V4, whereby you could do as you say, and best fit, (smooth) you could do a standard spline, OR, you could do a spline with emphasis on tangency, without curvature continuity.

To the best of my knowledge, such an option does NOT exist in V5.

Incidentally, this was one of my very first V5 questions.




**************
Check out CATBlog!

 

[itsmyjob]

Hi,

when you create a spline in V5 you pass thru the point and get a bumpy curve.

if you have FreeStyle you can create a 3D curve and chose near point option, you can then control the deviation of the curve from the points. FSS license is required.

if you do not have this license...

this is what i do :

if i have 20 points, i will create 2 splines in GSD from point 1 to 10 and from 10 to 20.
then i will join those 2 splines and i will have a tangency problem in my join (this is important).
then i will smooth the join.

the result is not exactly the ARC or the SPLINE function in V4 but the result is close to it.

Eric N.
indocti discant et ament meminisse periti ​

 

[Kapitan]

-Jackk,

Without knowing how many points are in your series, or seeing the overall trend of them, I would ignore, or delete, a particular point in an area that you don't like, and see what happens.

In V4 Arc, a result could be obtained that did not actually contact any of the given points, an extreme case being, to select every point and then set the (Polynomial) degree to 1. This gave, obviously, a line that was identical to a V4 Mean line, probably without touching any point.

Points 1,2,3,4,5,6,7 Points 1,2,4,5,7 Points 1,4,5,7

In V5, I've taken 7 arbitary points and fitted splines to them, using no tangency or curvature inputs, just point fitting. The first fits every point and gives a horrible looking curve, the second is an improvement - using 5 points, but to me, the best is a 4-point fit (cubic). Only the user can know the amount of allowable deviation. An exact fit of every point is far less likely to give a smooth curve because it is tied to too many points.

In V4, the same principle of defining a degree (1 - 15) applied in Patch for u and v directions. In some ways V4 was too clever, for day to day modelling and drafting, and a lot of methods have been streamlined when V5 was put together. Another way of emulating the V4 Arc fitting methods is to write some code and do it yourself. I normally do this sort of thing in Excel using a VBA macro. For a series of points an algorithm can be written to find the co-efficients of the polynomial for the fit required, and once the function is known, as many points as needed can be calculated for y = f(x) and a curve created in V5 directly from an external file.

But it is quite likely that V5 will give a smooth curve and not require any programming, if some start and finish tangency and maybe curavture information can be used, then tensions can be used and the curve manipulated that way.

 

[jackk]

Thank you to everyone who responded, especially Eric!

To be more specific, say I have 100 points that define an airfoil. And my point coordinates have been rounded off to 4 decimal places.

When I create a spline and analyze it, I see many inflections. I'm looking for a method to best-fit a curve through all these points (and avoid a curve that has 99 cells)

Eric: The Freestyle 3D CURVE-NEAR appears to do what I'm looking for. If you have any experience with it, I'd like to ask your advise on the various options that are available. Should I use a large or small deviation? (I'm thinking large) Does segmentation control how many cells? Which smoothing options work best?

thanks, Jack

 

[itsmyjob]

Hi,

yes segmentation control cells. catia will give you the segmentation N:6 ,(SEG:2) -> 2 cells each of order 5 (6-1)

small or large deviation.. this is up to you, usually i make a curvature analysis on the curve, this will help. if values had been rounded to 0.000x then you should not go above 0.00005 but again it is up to you.

Eric N.
indocti discant et ament meminisse periti ​