Bulges on a cylinder

[[Da...]]

Hi guys,

is there a way to get the LSQ (Gauss) diameter of bulges? I mean, I could do a 360° scan and choose max inscribed, but I'm pretty sure that wouldn't be the same.

See drawing and 3D model attached. The part is molded plastic, units are in mm.

inner_dia.zipinner_dia.pdf

[[No...]]

You can try to scan over each bulge on its own with a very high point density (circle segment, with restricted center if necessary) and get the min/max coordinate. Then recall all min/max points to a circle. Use a rotational pattern if possible.
We have parts like this quite often and that's how we usually do it when Gauss is required.

[[Da...]]

For fun, you could do a 360 degree circle and set the outlier sigma to 1 for inside part. this will eliminate most of the larger diameter sections. Then ask for max inscribed. You'll have to play around with the OL settings, but you can eliminate as much or as few as you want.

[[No...]]

Yes, I used to do that too for a while, but besides it being a fiddly procedure to find the right parameters (you often need to include a number of neighboring outliers as well), it turned out that even a slight deviation from the perfect settings could lead to unwanted points being left over. It depends a lot on the actual geometry.

[[Da...]]

Please sign in to view this quote.

Thank you, will try that when I find some time.

[[Da...]]

OK, with some help via PM I have it. I have attached printscreens of actual part.
I have scanned a line over bulge (took rotational pattern actually) with small speed and took quite a lot of points.
Then I took Maximum Coordinates/Points for each section of the line. First Max is in the Base Alignment, then Alignment1 follows and so on. I have entered X, Y and Z for Nominal in each Max point and recalled all Max Points into a new circle.

Should I also type in 0, 0, 1 for i, j, k vector to make it yellow?

Comments please 🙂

rec.PNGmax20.PNGmax19.PNGalign1.PNGline-str.PNGline-pat.PNG

[[An...]]

Please sign in to view this quote.

If you say Maximum Coordinate then take Maximum Coordinate!
Watch the constraints!

[[Da...]]

Please sign in to view this quote.

I'm not sure how to constrain angles.

Maybe this Normal Vector checked?

normal-vector-on.PNG

[[An...]]

That's right.
According my experience it is very important to understand the meaning of CONSTRAINTS
especially when using Min.-Max. Coordinate.
Do a research work here or elsewhere in order to improve your personal knowledge.

[[No...]]

A prerequisite for unsing lines here is that you must be absolutely sure about the positions of the bulges in relation to your alignment. The reason for this is that the evaluation direction for the mix/max coordinate is perpendicular to the line, so the nominal line must run perpendicular to the bulge.
I'm not sure whether or not constraining the angles is a must. But you must set the correct nominal angles and you absolutely must select the right reference element (nominal / actual) in the min/max dialog. In this case you want the nominal line for reference, because the actual line could end up under unpredictable angles, resulting in a wrong evaluation direction.

If you're in doubt, try circle segments with constrained center coordinates (and use the actual center of the circle as nominals), because for a circle the min/max is calculated in radial direction.

[[Da...]]

OK, it's been a while.
Note I'm only using 6 out of 8 max points for lsq circle.
Thank you for all suggestions and more comments please

The latest screenshots:

line-strat.PNGline-pattern.PNGline-eval.PNGalign1.PNGalign2.PNGmax1.PNGmax2.PNGcircle-strat.PNGcircle-eval.PNG

[[An...]]

Please sign in to view this quote.

One 360° circle probing is enough.
High point density is requested.

[[Da...]]

Thank you Andreas and Norbert, I will try that too (hopefully this year 🙂

[[Öz...]]

Hello Dane,

What kind of sensor do you have? If its a active sensor can you try to measure the cylinder with a disc probe and single points strategy. And you will rotate the points to bulges as much as you can on the CAD. Then disc probe will touch the edges of the bulges. Then you can evaluate with LSQ directly from the edges of the bulges. If you have a passive sensor you can try the same with a bigger probe. But disc probe will be more accurate.

[[Da...]]

I have Vast XT gold. Thanks, I will try your suggestion also for comparison.

[[Da...]]

Please sign in to view this quote.

I understand that part. But I haven't tried to compare what difference will it make in my case, since I would first like to go with the second method you suggest.

Please sign in to view this quote.

I did tried so, but again I'm having a hard time understanding constrains. Should I constrain the circle which has segments or circle in which I'm recalling the Minimum points? Or both? Should I also constrain radius on the one with segments?
See attached

segm.PNGminp.PNGrecall.PNG

[[No...]]

Constrain the cicle segments that you actually measure. You constrain when you want to make sure something stays at a fixed value resp. position. If you don't constrain the center, Calypso would try to freely calculate it from the shriveled little point cloud you scanned on each bulge. On such a narrow segment every small piece of dirt, a form deviation or a couple of points scanned beyond the actual edge could result in huge deviations of the center and thus of the calculated radius. So you need to tell Calypso where you expect the center to be. That's why you constrain here.

Later, when you already have all the minimum points, you no longer need to constrain, because then you have a 360 degree circle. Only if all the bulges together would still form a rather narrow circle segment, constraints might be an option.

[[Da...]]

Thanks Norbert, I get it now, at least for this situation. I'm bloating this topic pretty hard with printscreens, I guess I'm more a visual type.

constr1.PNGmin1.PNGt-cir-lsq.PNG

[[Da...]]

I also tried both, actual and nominal feature for reference in minimal points.
When nominal is used, diameter gets bigger, but only for 0.2 microns of mm.

Capture.PNG

[[Da...]]

Please sign in to view this quote.

Andreas, I'm sure your idea is pointing to the right direction. But, when I do circle segments I can choose circle1(1, 2, ...) in minimal coordinate points. I'm not sure how to go with one 360° circle, where all I can select is circle1 for example.

[[No...]]

Please sign in to view this quote.

If you use circle segments for min/max calculation and have the center constrained, the difference between nominal and actual as the reference should be near zero. Ideally, it should be exactly zero, but maybe the different radii cause tiny projection errors or something like that, so that sometimes a different point is identified as the minimum.