Curve Best Fit Alignments/Alignments from Several Curves

[[Jo...]]

Cross-Posting from the Curve forum in search of answers...

I recently completed my Curve training and the one thing that I left confused by, was the alignment options for individual curves. And maybe this just stems back to some sort of misunderstanding about best fit alignments to begin with. My vaguest memory of covering any best fit alignments in all my training is that it's mostly only when stuff is not fully constrained. Someone please correct me if I am wrong on that account.

Going back to alignment of curves; are these options only intended for curves that are not fully constrained by a datum structure? An example would be a curve(s) dropped onto a spline, on some outside geometry that lies perpendicular to a plane. In the example the plane would be a datum, and the splines would be used in a profile callout only called back to that one plane, which would obviously not be fully constrained.

The real world problem I have is we have several new parts we are making that I am programming, with splines that define the outside geometry on. The parts are mostly "circular-shaped" but have several pieces of different geometry (planes, huge radii with small radii and tiny arc segments that transition onto other geometry). In one example, I have a .002 profile back to A-B-C, which this datum structure does fully constrains the part. Does this render a curve best fit redundant since it's fully constrained? Just out of curiosity I did create curve best fit alignments, then alignment from several curves, and they did improve the results which makes me feel.... uneasy.

To put my method simply: I created curve best fit alignments, used those to create a multiple curve best fit, created a secondary alignment using A-B-C and referenced the multiple curve best fit. It improved the results... but I feel like I'm doing the equivalent of forcing a gage pin through a hole that barely fits just so I can say "the gage pin goes".

[[Ma...]]

Jesus - what a bunch of a text 😱

If you have fully set datums ( A-B-C ) then you can not use alignment from curves ( why would you - you must uncheck every translation/rotation ).

Any better articulated question?

[[Jo...]]

Please sign in to view this quote.

Thanks, that was very useful.... sorry you had to read a couple of paragraphs to find a way to give a snippy response. Next time I'll just post "How to do (insert question)???"
You did sort address my question, but best fit alignments are something we have not had much need for in the past, for the work we do. I am trying to get feedback from people who have. Thanks.

[[Ma...]]

Curves rely on nominal data - can be used for pre-alignment of precise scans - ex. if you have bended parts.

We are using it for part where is involute profile and plane have perpendicularity to it.

[[To...]]

Please sign in to view this quote.

I am currently measuring 5 curves on a blade. Then recalling those 5 scans into a Best Fit from Multiple Curves (6 degrees of freedom). Then I measure 5 new curves (copies of originals) using the previous curves best fit. I repeat this a total of 4 times then analyze Line Profiles using the last curves best fit. Overkill? Maybe.

curves best fit 1.JPGcurves best fit 2.JPG

[[Ow...]]

In general, no, curve best fit's should not be used but, there are always exceptions.
Curve best fits move the actuals to align with the nominal path in whatever axis you choose so, because you are moving the actuals, you have to be careful to ensure you're not making something bad look better than it is or vice versa.

As important as anything is to ensure you have your filters set correctly or masked to ensure you're not getting any bogus data collected at the start/end of a curve or when the curve transitions across a sharp edge or inside a radius smaller than the probe radius or a corner.

I don't use the profile characteristic much in our application so, I don't have a lot experience with it but, how you evaluate it is also critical, examples- bilaterial, one result, two results, unilaterial, outward to infinity, etc, it's stuff not always defined correctly on drawings. I've attached a good example of how the evaluation can change the outcome.

Without a drawing or plot examples, it's difficult to determine what's best and your best option is probably to call the help line (you just took a class so it should be free if you don't have an active software maintenance agreement) and have them log-in via on-line to where they can see what you're working on and get you going in the right direction.

Regarding the snippy reply, it's not always the case but, keep in mind that people reply on here from countries all around the world that speak different languages and a google translation doesn't always get the words right and that what might offend somebody in one culture is just fine in others.

Profile, Curve Form and Profile Calypso.pdf