sufficient datum size?

[[Re...]]

Suggestions – What measuring device is accurate enough to measure runout on an OD cylinder, which is 4.50” away from the primary cylinder datum that is less than 0.75” long, to a tolerance of 0.001”? Both cylinders share the same center-line and both have grind finishes. The diameters of both cylinders are 0.75”. I have some parts that have failed on an Accura using a Vast head. I feel like I have a fairly robust program. However rerunning these parts at different dates with a fresh calibration and having slightly different environmental conditions, such as temperature, airflow, humidity, vibration and so on. I get variation in this one particular runout characteristic of up to 0.0005”, but all other characteristics in the program seem to repeat, including a number of other runout characteristics with the same tolerance, but using larger datums. I believe that this repeatability error is caused by a projection error mainly due to the primary datum not being sufficient in size. However I’m not sure if anybody is buying this reason, so in the meantime with confidence being lost with the cmm by many, I need to find an alternate method to measure these parts. I personally cannot see an alternate method without going to a device that is more accurate than the Accura cmm. Are there any zero spindle gauges out there accurate enough? Any thoughts?

[[Ch...]]

Projection error is your problem. See strategy that starts on page 89 of this slide deck. The whole thing is worth reading though.

Profile_Concentricity_CALYPSO.pdf

[[Da...]]

Work holding has to be solid, super solid. Put the part in position as you would run the part...then set up a tenth indicator touching the second cylinder and poke the part lightly, if the indicator moves at all, it's not going to work. Loop your alignment, use proper filtering and outlier elimination.

[[Re...]]

Chad,
I totally agree with you. The presentation of using A-B as a datum callout is the method that I have convinced most of our designers to use. The part that I'm having issues with uses this type of callout on all of the gd&t characteristics except the one in question, which is runout of B to A. No ability to change this print, it is not our design, it's a large aerospace company. So I need to find a more reliable way to satisfy their measurement callout.

[[Re...]]

Dave,
I agree with all of your points. Especially the looping of the alignment. I feel that so many novice programmers overlook the importance of this step. I know that it was not taught in my training courses many many years ago. I had to learned a lot the hard way.

[[Da...]]

Please sign in to view this quote.

I've been working with ultra-close tolerances for 6 years. I learned how to deal with these problems. the train drivers don't always understand. But really, work holding is a major player. The further the projection, the tighter everything has to be. Were the two diameters ground at the same time? (Centerless? between centers?)

[[Re...]]

Dave,

The 2 diameters were ground between centers but not at the same time.

[[Ch...]]

Explain it with pics like this. Even micron level measurement error on the primary datum can have a decent impact.

[[Ri...]]

In the past I have measured each cylinder as two sperate circles per cylinder.
Then check runout of each to each other. You end up with 4 runouts, then just use the max value.
I eliminate A1 and A2 errors. I had a white paper once from a seminar, it was called:
Short Unstable Datums or Features.
This did not mean the part was moving, it means there are inherent errors in the A1 and A2 values of the features.
The paper was written in early 2000's.

[[Re...]]

I understand the datum projection thing. The part that I'm having a hard time understanding is my maximum possible measurement error on a cylinder 0.75" diameter x 0.75" long taking 2 circles scans. My machine MPE E = 1.9 + L / 300. If I was looking at a single point measurement in 1 axis, I think that I may understand. But the 3-D axis error potential on the cylinder with 2 circle scans, is what I do not comprehend. Once I figure out how to calculate this for the cylinder, the projection error is then easy to calculate. From there, maybe I could work backwards to show how accurate of a machine is needed or how much tolerance is needed for my characteristic to stand a chance with an R&R study.

[[Ke...]]

Please sign in to view this quote.

I'm no measurement uncertainty guru... but I would start with "What is the highest allowable form deviation for the feature", and then presume that each measured circle has a maximum allowable deviation at opposing points (e.g. Circle 1 deviation is in the +Y direction, and Circle 2 deviation is in the -Y direction), and calculate how that may move the circles center point (e.g. Circle 1 = (+0.001) and Circle 2= (-0.001)) from a circle with theoretically perfect form, and then calculate the highest possible variation, given that the datum form is within spec.
Personally, I always include a form characteristic for all datums (in this case I would include roundness for each circle path, as well as cylindricity), even is one is not explicitly specified.

[[Ow...]]

If you could share the drawing or a part of it with proprietary info hidden, that would be helpful.
Not knowing what the part is or how it's intended to be used makes it impossible to determine if the the drawing specs are correct to begin with.
However, generally you want the length of a cylinder to be at least 5x the size of the diameter for any cylinder to measure well enough to project anything to it.
If you're measuring both cylinders with two scans, you might take one scan at the farthest end of one and take the other at its farthest away scanned position and recall them into a 3D line, then use that 3D line as a Datum and report the XYZ deviations of the other sections of the scan to get an idea of how they align.
Or recall both features into a stepped cylinder (helps constrain it) and report the diameter, XY or Z deviations of each section. I think you can use a cumulative runout characteristic with the stepped cylinder and it should be much more repeatable.

[[Re...]]

So I could have a +1.9 µm deviation at 1st circle scan and a -1.9635 µm at 2nd circle scan. This is looking at it in a 2D world, I believe that the error could be more if looked at in 3D. But for simplicity lets say that I could have a 4 µm difference in the position of each circle scan looking at just one view, then the axis of the cylinder could be tilted by up to 0.01203°. This would reflect a projection error of over 0.0009" at 4.5". Unless I'm missing something, I would have to say that I'm very lucky that my cmm calibration is much better than it's allowable limit, or I would not even be able to come closed to measuring the runout characteristic in question.

MPE E.gif

[[Re...]]

Runouts to Datum A-B are no problem. There is about a 4.5" spread between the cylinders and the cylinders are approximately .75" in diameter. It's the runout of B to Datum A that is hard to control. Is there better way to have to have this feature controlled? If not,should I push for more tolerance?

part.JPG

[[Ch...]]

So what have we learned? I would say we learned that it's not the size of your datum that is important, but how you use it. 🤣 🤣

[[Er...]]

Maybe propose you check Datum B to A-B runout within .001" as well as check Datum A to A-B runout .001". It seems the application would be the two Datums working together.

[[Ke...]]

If the two datums work together, and have the same size and form callouts, then they essentially act as a single (interrupted) feature... I would think that they should be evaluated as a single cylindrical feature (Cylinder A-B), and that Cylinder A-B should meet the print size and form requirements ❓

[[Re...]]

The cylinders are slightly different in size, but have the same tolerance. No roundness callouts, just runout characteristics.

[[Re...]]

When I see this situation with projection error, I'm able to get the datum scheme changed on internal prints. I don't have much pull on getting a customer prints changed. However with your inputs, I stand a better chance.
Thanks for everyone's comments.

[[Ja...]]

If A and B were the same diameter you could scan them as a single cylinder and use cad evaluation to see where and why the runout will fail. Maybe you could do the same with a step cylinder but I have never used a step cylinder myself.

[[Ke...]]

I would also revisit the basics...
What is the repeatability of the measurement for a given part? (high repeatability will increase confidence in measurement results)
Use a slow measurement speed, for better repeatability.
What is the actual form of the datums (especially Datum A) - the form for short datums need to be very tightly toleranced.
I have also had good luck with increasing measurement locations and point density, to stabilize (as much as it can be) a short datum... say using 5-7 scan paths, instead of 2.
Sort through a handfull of parts to find the best/worst case results, to try and determine the root cause of the difference in results.

[[Re...]]

The allowable accuracy error in the cmm alone can induce 0.0009" projection error with this short of datum. My repeatability for my runout measurement will vary up to 0.0002" without moving the part. But I have had a variation of up to 0.0005" with a reload, but only one time that I know of. I would say that my part is stable, or else my other callouts that use A-B as the primary datum would have repeatability issues also. My form error on these cylinders are 0.0001" at worst case. Parts have been measuring in tolerance on this particular callout for quite sometime. I had a part measure out of tolerance, and weeks later ask to measure again. At that point the repeatability was found to be failing. Know I need to show why the cmm is failing. If I can get the primary Datum to become A-B for all of my characteristics, then all my problems go away. I say that it is projection error, but I need to convince others.

[[Re...]]

FYI - I do use 2 thermal-couplers for temperature compensation with my plans.

[[Ma...]]

If someone argues against evaluating runout A and B to the compound datum A-B you should ask them to sign their name with a pencil but only allow them to hold the last half inch by the eraser when they do it. Maybe they'll catch on to the problem. 😃

[[Re...]]

Mark, LMAO