Concentricity of Sphere to Cylinder

[[Ja...]]

I have a part with an I.D. sphere and an O.D. cylinder. I align the part and then take some scans on the sphere to check concentricity (sphere to cylinder). I am evaluating all features LSQ. Every part is reading bad, consistently, for concentricity. When I change the sphere evaluation to inner tangential element, I get one good measurement. Before I modify the program and say that this is the correct way to evaluate the sphere, any opinions?

[[Cl...]]

How is the part aligned?

[[Ja...]]

I set the part on the table, level and origin Z to the table. Origin X and Y on the cylindrical O.D.

[[Cl...]]

So you use the cylinder for the spatial rotation & Z origin?

[[Ja...]]

The part has a flange at the base. I measure it as a plane and that is where I do the spatial rotation (+Z) and Z origin. The O.D. cylinder just gets the X, Y origin. The sphere is not part of the alignment. There is no planar rotation.

[[Cl...]]

If the cylinders length is at least 3x it's own diameter, Try spatial rotating the cylinder in Z, and see what happens.

[[Ja...]]

That isn't the case here, the diameter is roughly 4" and the height is about 3".

[[Br...]]

How are you measuring the Datum Axis? Helical scans can cause problems when establishing Datum Axis'. Also, how are you measuring the circles on the Sphere? Are you using the "circle on a sphere" feature to scan it? If not your vectors will get pretty funky as you approach the poles. Unfortunately Calypso doesn't actually evaluate Concentricity correctly. It treats it the same as Position. That being said I don't think Inner Tangential is the correct method for this. Its going to take all the inner most points to derive a center point, or axis from, wheres it should always come from the most outer to establish the actual mating envelope. But since what you really want is concentricity and not Position, Least Squares is probibly closer to the median points that should be evaluated for that geometric tolerance.

[[Ja...]]

I have attached an image that may make the situation more clear. For the O.D. cylinder (datum), I am measuring with 2 (or 3) 360deg circle paths. For the sphere, I am using 8 circle paths. I come in at 45deg with the probe, and run 4 scans (roughly 60deg travel) along the sphere. Come up, rotate, and measure a second side of the sphere with similar paths.

The results for Concentricity and Position are identical in Calypso, so it is treating them the same. I just noticed that when I switch to Inner Tangential, the measurement reads good. I'd like to know more about which method is appropriate, but with what you're saying LSQ may be better, thus making every part bad.

[[Br...]]

Yeah Concentricity is really hard to evaluate, but also its misapplied 99.9% of the time. I'm sure Position is actually what is wanted here for the design. Concentricity has almost no use. In fact, I believe its being dropped in the next revision of Y14.5. If you want to evaluate it as a position tolerance, you should either use Outer Tangential, or Maximum Inscribed as a fitting algorithm. This will create the actual mating envelope to derive the center point of that sphere. Also, that cylinder datum feature needs to be either Outer Tangential or Minimum Circumscribed regardless of whether its being evaluated as position or concentricity. One more questions: Are you filtering and eliminating outliers?

[[Ja...]]

Oh wow, I didn't know Concentricity was so useless that it may be thrown out.

Is it correct to think that the logic behind the choice of fitting algorithm must be related to the consideration of the functionality of the part? For example, you say that the cylindrical datum must be either OT or MC. Is this because it is an O.D. and an algorithm like LSQ may calculate the center incorrectly, thus making it possible to accept OOT parts? I apologize for my lack of GD&T understanding.

I am Gauss filtering and removing anything over +/- 3 sigma.

[[Br...]]

Right, what we want to achieve with the algorithms is what best represents what our standard (Y14.5) says. So in the case of a position tolerance Y14.5 states that Position is the location of one or more features of size relative to one another or to one or more datums. A positional tolerance defines either a zone within which the center, axis, or center plane of a feature of size is permitted to vary from a true (theoretically exact) position.

So now that we know that we have to ask ourselves, how is a center, axis, or center plane derived from a feature of size? Well once again, Y14.5 says this is done with whats called an actual mating envelope. This envelope is outside the material. A similar perfect feature(s) counterpart of smallest size that can be contracted about an external feature(s) or largest size that can be expanded within an internal feature(s) so that it coincides with the surface(s) at the highest points.

In other words, what you have with your application, is an OD cylinder that is a feature of size and is your primary datum. To use its axis as a datum, you must have a cylinder that contracts around your actual part until it contacts all the high points. Minimum circumscribed, or Outer Tangential will do exactly that. With your sphere you have a sphere that is expanding until it contacts the high points to derive a center point of the sphere. Maximum Inscribed or Outer Tangential will achieve this.

Now there's a lot more to be said about these concepts regarding the difference between Related and Unrelated actual mating envelopes, but before I confuse you too much, I'll leave it at that for now.

[[Ja...]]

Thank you Brett,

Your explanation makes this issue much easier to understand. I will modify the program accordingly and try to think more along these lines going forward.

Thanks